A Corpus Based Analysis of 700 Historical Wills
Across Six Jurisdictions, 1873 to 1967
Aphiwat Bualoi
Independent Legal Educator, Chiang Mai, Thailand
ABSTRACT
This article reports findings from the 1,000 Wills Project, a corpus based investigation of approximately 700 historical testamentary instruments drawn from ten archival collections spanning the United States, Canada, Australia, and New Zealand. The corpus covers the period from 1873 to 1967 and comprises 1,313 scanned document images. Using a multimodal artificial intelligence system capable of reading handwritten and typewritten archival documents, the author conducted systematic structural analysis across the full corpus, supplemented by detailed close reading of seventy representative instruments. The study identifies nine universal structural elements present in virtually all wills regardless of jurisdiction or period, documents the decline of religious invocatory language between 1920 and 1935, establishes the spousal life estate as the dominant distribution pattern across all six jurisdictions, catalogues seven recurring drafting failures, and proposes a universal structural template grounded in empirical evidence rather than doctrinal prescription. The article contributes to the fields of comparative succession law, legal history, and legal pedagogy, and demonstrates a methodology for artificial intelligence assisted archival legal research.
Keywords: wills, testamentary instruments, corpus analysis, comparative law, succession, legal history, artificial intelligence, probate records, estate planning, legal pedagogy
I. INTRODUCTION
The study of testamentary instruments has long relied on doctrinal analysis of appellate decisions and statutory commentary. The established treatises present will drafting as a deductive exercise in which the practitioner learns the governing rules, studies model clauses, and applies them to the circumstances of each client. This approach is doctrinally rigorous, yet it suffers from a notable empirical gap. It instructs students in what wills ought to contain according to law, but it does not instruct them in what wills actually contain in practice, how drafting conventions evolve over the course of decades, or how structural patterns remain stable or vary across jurisdictions.
This article seeks to address that gap. It reports findings from the 1,000 Wills Project, a corpus based investigation of approximately 700 historical testamentary instruments spanning six countries, ten archival collections, and nearly a century of legal practice. The corpus comprises wills probated between 1873 and 1967 in the United States, Canada, Australia, and New Zealand. The study examines these instruments not as isolated legal documents but as a population, employing quantitative structural analysis to identify patterns that no examination of a single document could reveal.
The research was motivated by a pedagogical concern. The author teaches will drafting in Thailand, where students must acquire familiarity with both the Thai Civil and Commercial Code, Book VI on Succession (Sections 1599 to 1755), and the common law testamentary conventions that govern international estate practice. The instructional materials currently available offer doctrinal guidance but lack the empirical foundation that would allow students to observe how real wills are structured, how conventions change over time, and what drafting failures recur across jurisdictions. The 1,000 Wills Project was conceived to supply that foundation.
A central methodological innovation of this study is the deployment of a multimodal artificial intelligence system to read and analyse scanned archival documents at scale. Traditional optical character recognition is inadequate for historical handwritten and typewritten legal documents, which exhibit inconsistent letterforms, faded ink, irregular orthography, and overlapping page layouts. The multimodal system employed in this study reads document images directly and extracts not merely textual content but legal structure, identifying clause types, bequest patterns, and drafting omissions with a degree of comprehension that earlier technologies could not achieve.
The article proceeds as follows. Part II describes the corpus and methodology. Part III presents findings in six areas, namely universal structural elements, religious and mortality language, spousal life estate patterns, disinheritance techniques, testamentary migration evidence, and recurring drafting failures. Part IV discusses the implications for legal pedagogy and comparative succession scholarship. Part V offers concluding observations.
II. CORPUS AND METHODOLOGY
A. Composition of the Corpus
The corpus consists of approximately 700 testamentary instruments drawn from ten archival collections. The collections were selected to maximise geographic, temporal, and socioeconomic diversity within the common law tradition. Table 1 summarises the composition of the corpus.
Table 1. Corpus Composition by Collection
| Collection | Period | n (approx.) | Pages |
| New York Probate Records | 1629 to 1971 | 182 | 385 |
| Maryland Register of Wills | 1629 to 1999 | 100 | 167 |
| Illinois Probate Records | 1819 to 1988 | 100 | 103 |
| West Virginia Will Books | 1756 to 1971 | 100 | 132 |
| Canada, Manitoba | 1871 to 1930 | 50 | 102 |
| Canada, British Columbia | 1861 to 1981 | 53 | 79 |
| Canada, Saskatchewan | 1887 to 1931 | 51 | 71 |
| Canada, Nova Scotia | 1760 to 1993 | 50 | 61 |
| Australia, Victoria | 1841 to 1926 | 12 | 25 |
| New Zealand | 1843 to 1998 | 100 | 188 |
| Total | approx. 700 | 1,313 |
Each instrument in the corpus exists as one or more scanned images in JPEG or PNG format, photographed from original probate records, will books, or court filings. The images range from single page holographic wills to multi page solicitor drafted trust instruments. The testators span a wide socioeconomic range, from illiterate farmers who signed by mark to retired solicitors who constructed complex testamentary trusts funded by publicly traded securities.
B. Analytical Method
The study employed a two stage methodology. In the first stage, all instruments in the corpus were surveyed using artificial intelligence assisted image analysis to extract basic structural data: the presence or absence of standard clause types (preamble, revocation, debts, specific bequests, residuary clause, trust provisions, executor appointment, testimonium, and attestation), as well as executor type, bequest types, and attestation format.
In the second stage, seventy instruments were selected for detailed close reading using a seven point analytical framework developed by the author. The framework evaluates each will across seven dimensions: intent and purpose, structure and organisation, language and precision, bequests and distribution, executor and administration, safeguards against contest, and modern fitness. Each dimension is scored on a scale of one to five. The seventy instruments were selected to ensure representation of all ten collections, all major time periods, and a range of estate complexities.
The artificial intelligence system was deployed as a reading instrument, not as an analytical one. All pattern identification, all interpretive judgment, all pedagogical selection, and all template construction were performed by the author. The role of the system was confined to image to text conversion with structural tagging, a task that, for 1,313 archival images, would have required several years of manual effort by a single researcher.
III. FINDINGS
A. Universal Structural Elements
The most significant finding of the study is the near universal consistency of clause order across the full corpus. Nine structural elements appear in a fixed sequence in virtually every instrument examined, regardless of jurisdiction, time period, or estate complexity. Table 2 reports the observed frequency of each element.
Table 2. Frequency of Structural Elements Across the Corpus
| # | Element | Frequency | Classification |
| 1 | Preamble (identification and capacity declaration) | 100% | Universal |
| 2 | Revocation of prior instruments | 97% | Universal |
| 3 | Direction to pay debts and funeral expenses | 99% | Universal |
| 4 | Specific bequests | 65% | Variable |
| 5 | Residuary clause | 92% | Near universal |
| 6 | Trust or life estate provisions | 45% | Variable |
| 7 | Executor appointment | 100% | Universal |
| 8 | Testimonium (signature and date) | 95% | Universal |
| 9 | Attestation clause (witness certification) | 100% | Universal |
Six of the nine elements (the preamble, the revocation, the debts provision, the executor appointment, the testimonium, and the attestation) constitute the minimum viable will across all jurisdictions represented in the corpus. Their sequence is invariant. No instrument in the corpus places the executor appointment before the bequests, nor the revocation clause after the debts provision. This structural invariance holds equally across handwritten holographic wills, pre printed probate court forms, and solicitor drafted trust instruments.
The remaining three elements (specific bequests, residuary clause, and trust provisions) correlate strongly with estate complexity. Modest estates, such as those of a schoolteacher or a retired gardener, typically produce instruments containing a single universal bequest to the surviving spouse. Substantial estates, such as those involving portfolios of publicly traded securities or extensive agricultural landholdings, produce multi page instruments containing multiple bequest types, trust structures, and detailed executor powers.
B. Religious and Mortality Language
The corpus reveals a datable transition from religious to secular preamble language. The invocatory formula, ‘In the name of God, Amen,’ which descends from medieval English ecclesiastical testaments, appears in ninety three per cent of Illinois instruments dated before 1920 and in eighty five per cent of New York instruments dated before 1900. The formula is entirely absent from all instruments in the corpus dated after 1935.
The transition proceeds through an identifiable intermediate stage. Five Illinois instruments from the period 1912 to 1920 employ mortality awareness language without religious invocation, using phrases such as ‘considering the uncertainty of this frail and transitory life.’ This formulation acknowledges mortality in quasi philosophical rather than theological terms. By the mid 1920s, even this language disappears, supplanted by the purely functional declaration of ‘being of sound and disposing mind and memory,’ a formula whose purpose is to establish testamentary capacity in the event of a contest rather than to express spiritual conviction.
A notable exception exists within the French Canadian notarial tradition. The 1891 will of Zephirin Roy, drafted before a notary in the Parish of St. Patrice de Beaurivage in the Province of Quebec, opens with the words: ‘I give my soul to God, begging Him to give it His mercy.’ This formulation, which is distinct from the English language invocatio, reflects the civil law notarial tradition and persists considerably later than its common law counterpart. The will was subsequently probated in Saskatchewan, having followed the testator westward during the great French Canadian settlement of the Prairie provinces.
The decline of religious invocatory language should not be interpreted as evidence of declining religious practice among testators. Rather, it reflects a shift in the professional standards of the drafting bar. The older invocation served a spiritual function for the testator. The modern capacity declaration serves a legal function for the protection of the estate. Practitioners replaced one with the other not because the spiritual dimension ceased to matter to their clients, but because the capacity declaration acquired greater evidentiary weight in contested proceedings.
C. The Spousal Life Estate as Dominant Distribution Pattern
Approximately forty five per cent of all instruments in the corpus employ a spousal life estate as the primary mechanism of distribution. Under this arrangement, the surviving spouse receives income from, or use of, the estate during her lifetime, and the corpus passes to children or other blood relatives upon her death. This pattern appears across all six jurisdictions and all time periods represented in the corpus.
Representative examples include instruments from Illinois (1896), New York (1909), New Zealand (1942), and New Zealand (1964). The language varies from informal constructions, such as ‘the use of all property during her natural life,’ to formal trust language, such as ‘upon trust, to pay the net annual income to my wife during her lifetime.’ The economic structure, however, is identical in every case. The surviving spouse receives a usufructuary interest. The underlying capital is preserved for transmission to the next generation.
A subset of life estate instruments, comprising approximately ten per cent of those employing the pattern, include a clause terminating the wife’s interest upon remarriage. The concern motivating this provision, which is confirmed by the language of the instruments themselves, is that a second husband would absorb the assets of the family estate. This provision reflects a social norm regarding the economic dependency of women that was legally operative throughout the period covered by the corpus but has largely fallen out of favour in modern drafting practice across the jurisdictions examined.
D. Disinheritance Techniques
The corpus documents two distinct techniques for effecting disinheritance. The first and more prevalent technique is the nominal bequest, in which a token sum, typically between one and five dollars, is given to a child whom the testator intends to exclude from any meaningful share of the estate. Multiple instruments from Illinois, British Columbia, and New York employ this device. The nominal bequest functions as a legal prophylactic. It demonstrates that the testator was aware of the child’s existence and deliberately chose to provide only minimal provision, thereby defeating a potential claim by a pretermitted heir.
The second technique, observed in a single instrument but of considerable analytical significance, is the reasoned explanation. In a 1915 New York instrument, the testator distributes unequally among his daughters but includes an explicit statement of his rationale. He explains that the favoured daughter ‘has always lived and remained at home and cared for and looked after myself and my wife from day to day,’ whereas the other daughters ‘are well married and have been for many years caring for their own families and interests.’ This approach, which substitutes explanation for the token bequest, is increasingly favoured by modern drafting authorities as more resistant to contest and less susceptible to the implication of testamentary spite.
E. Testamentary Evidence of Geographic Migration
An unanticipated finding of the study is the considerable value of the corpus as a record of geographic migration. Multiple instruments reveal cross jurisdictional lives that are not readily visible in any other category of archival record. A will drafted before a Quebec notary and subsequently probated in Saskatchewan reflects the great French Canadian westward settlement of the Prairie provinces. An instrument executed in Galesburg, Illinois, and probated in Manitoba evidences Scandinavian American northward migration into western Canada. A resident of Ladysmith in British Columbia who retained ownership of one hundred acres of farmland in Cape Dauphin, Nova Scotia, illustrates the Maritime to West migration pattern that reshaped Canadian demographics in the late nineteenth and early twentieth centuries.
This finding carries implications that extend well beyond legal scholarship. Testamentary instruments, by virtue of their routine identification of places of residence, property locations, and family relationships, constitute an underutilised source for the study of migration history, historical demography, and the formation of diaspora communities.
F. Recurring Drafting Failures
The corpus permits the identification of drafting omissions that recur across jurisdictions and time periods with sufficient frequency to suggest systemic deficiencies in practitioner education rather than isolated lapses of individual judgment. Table 3 catalogues the seven most frequent failures observed during the detailed reading of seventy instruments.
Table 3. Recurring Drafting Failures in the Corpus
| Omission | Consequence | Observed Frequency |
| Absence of a predeceased spouse contingency | The estate passes by intestacy if the primary beneficiary predeceases the testator | Common across all collections in simple instruments |
| Absence of a successor executor | The court must appoint an administrator, increasing cost and inviting family disputes | Present in the majority of simple instruments |
| Absence of a residuary clause | Property not covered by specific bequests passes by intestacy | Approximately eight per cent of the corpus |
| Absence of a per stirpes provision | A lapsed bequest to a predeceased beneficiary is not redirected to that person’s issue | Approximately sixty per cent of instruments |
| Use of precatory rather than mandatory language | The direction is unenforceable because it reads as a wish rather than a command | Scattered across all collections |
| Absence of an executor powers clause | Court supervision is required for the sale of assets, causing delay and expense | Predominant in instruments dated before 1930 |
| Absence of a bond waiver for the executor | The estate bears the unnecessary cost of a surety bond | Approximately twenty per cent of instruments |
The most consequential of these failures is the absence of a predeceased spouse contingency. This omission is present in every instrument that employs the simple residuary to spouse pattern without specifying an alternative distribution. Should the named beneficiary predecease the testator, the residuary estate passes by intestacy, a result that is almost certainly contrary to the testator’s intent. The frequency with which this omission appears across jurisdictions and time periods strongly suggests a systemic gap in the education and training of drafting practitioners rather than an isolated error attributable to any single individual.
IV. DISCUSSION
A. Implications for Comparative Succession Scholarship
The structural invariance documented in Part III has significant implications for comparative law. The nine element sequence identified in this study is not prescribed by the statute of any single jurisdiction. Rather, it emerges from actual practice across multiple independent common law jurisdictions over a period of ninety four years. This observation suggests that the sequence reflects an inherent logic of testamentary disposition, proceeding through the stages of identification, revocation, clearance of debts, distribution of property, appointment of a fiduciary, and authentication, rather than a mere convention peculiar to any one legal system.
This finding lends support to the hypothesis that testamentary formalities, notwithstanding superficial variation in their statutory expression, converge upon a common structural grammar. The Thai Civil and Commercial Code, for example, prescribes attestation requirements that differ in important respects from those of the common law (Sections 1656 to 1660) and recognises forms of testamentary disposition, such as the oral will under Section 1663, that are unknown to common law jurisdictions. Yet the underlying structural logic of the Thai testamentary instrument, proceeding through identification, revocation, distribution, appointment, and authentication, is functionally identical to the sequence observed in this corpus. A structural model of the will that is grounded in empirical evidence from actual instruments, rather than derived from any single statutory regime, may therefore prove useful for comparative legal education well beyond the confines of the common law tradition.
B. Implications for Legal Pedagogy
The corpus based approach demonstrated in this study offers a model for legal education that differs from traditional doctrinal instruction in three important respects.
First, the approach is inductive rather than deductive. Rather than deriving drafting principles from abstract rules, students are invited to discover them from evidence. A student who reads thirty residuary clauses drawn from thirty instruments across four jurisdictions develops an intuitive understanding of the clause’s function, language, and permissible variations that no textbook description can replicate.
Second, the approach is comparative by design. Because the corpus includes instruments from the United States, Canada, Australia, and New Zealand, a student is naturally led to compare how the same legal concept, whether it be the attestation clause, the executor’s bond, or the spousal life estate, is expressed across four distinct common law traditions, and then to consider how it would be expressed under Thai civil law or the domestic law of any other jurisdiction.
Third, the approach embeds historical and human context within the framework of technical instruction. A student who reads a father’s 1915 explanation of why he favours the daughter who stayed home to care for him does not merely learn a drafting technique. A student who encounters a Quebec farmer’s 1891 commendation of his soul to God witnesses the intersection of faith and law in a manner that doctrinal analysis alone cannot convey. These encounters are not tangential to the purposes of legal education. They are central to it, because they remind students that the law of succession exists to carry out the wishes of real human beings who are confronting their own mortality.
C. The Role of Artificial Intelligence in Archival Legal Research
The use of a multimodal artificial intelligence system to read archival legal documents at scale constitutes a methodological contribution that is independent of the substantive findings reported above. The system employed in this study achieved three capabilities simultaneously: the conversion of handwritten and typewritten document images into structured text, including images that were faded or degraded; the identification and tagging of clause types within the extracted text; and the correct disambiguation of overlapping documents, such as two separate wills printed on adjacent pages of a single probate book.
It must be emphasised, however, that the system functioned as a reading instrument and not as an analytical one. All pattern identification, all interpretive judgment, all decisions regarding the construction of the template, and all pedagogical choices were made by the author. The appropriate analogy is the microscope. It permits the researcher to perceive what was previously invisible to the naked eye, but it does not perform the science. The researcher formulates the questions, designs the analytical framework, interprets the observed patterns, and draws the conclusions. The artificial intelligence reads the documents.
The practical significance of this capability is primarily one of scale. The careful reading of 1,313 archival images, many of them handwritten, many of them faded, and many of them in irregular formats, would have required several years of sustained manual effort by a single researcher working without institutional support. The artificial intelligence system completed the equivalent reading in a small fraction of that time, thereby enabling a scope of analysis that would otherwise have remained impractical.
V. CONCLUSION
This study has demonstrated that a corpus of 700 historical testamentary instruments, when subjected to systematic analysis with the assistance of artificial intelligence, reveals structural regularities of considerable stability and generality. Nine structural elements appear in a fixed sequence across six jurisdictions and ninety four years of practice. Religious preamble language undergoes a datable transition between 1920 and 1935. The spousal life estate dominates distribution patterns across the full corpus. The nominal bequest remains a widespread disinheritance device, though it may increasingly yield to the explanatory approach. Testamentary instruments constitute an underutilised source for the study of migration history. And a set of identifiable drafting failures recurs with sufficient frequency to suggest systemic deficiencies in the education of practitioners.
The universal structural template derived from this corpus, comprising a nine element sequence supported by empirically observed frequency data, is offered as a practical tool for comparative legal education. It provides a model of testamentary structure that is grounded in the evidence of actual instruments rather than in the prescriptions of any single statutory regime, and that may therefore be adapted to any domestic legal system, including civil law systems whose testamentary formalities differ in their specific requirements but converge in their underlying logic.
The 1,000 Wills Project remains an ongoing undertaking. Approximately 300 instruments in the corpus have not yet received the detailed close reading that the analytical framework requires. The template can be further refined as additional data becomes available. Future work will extend the analysis to these remaining instruments, explore the potential of the corpus as a source for historical demography, and examine the extent to which the structural model identified here may be applied to testamentary forms originating in civil law jurisdictions.
The wills in this corpus were written by real people. Farmers, widows, veterans, immigrants, labourers, and solicitors sat down, in a lawyer’s office or at a kitchen table, to record their final wishes. They concerned themselves with the welfare of their spouses, the futures of their children, and the maintenance of their graves. They attempted to be fair. They attempted to be wise. They did not always succeed. Reading 700 of their wills has taught the author more about the law of succession than any treatise could, not because the legal doctrine was more clearly explained, but because the wills revealed what the doctrine exists to accomplish. The law of wills exists to carry out the wishes of the dead. Every clause, every formality, and every rule of construction exists because someone, at some point in the long history of testamentary practice, wrote a will that failed, and the law evolved so that such a failure need not happen again.
About the Author
Aphiwat Bualoi is an independent legal educator based in Bangkok, Thailand. His research concentrates on comparative testamentary practice across common law and civil law jurisdictions, with particular attention to legal pedagogy for cross border estate planning. The 1,000 Wills Project is an ongoing research initiative. The Will Analysis Framework and the Universal Will Template developed in connection with this project are available as companion documents. Correspondence may be directed to bualoi.law@hotmail.com.
Aphiwat Bualoi Law Office 
Leave a Reply