Surname Spelling Variations: How to Search Historical Records

Why the name on the record is only one of many spellings — and a practical method for finding your family anyway

Every family historian eventually hits the wall. Grandad’s surname is Johnson, but he cannot be found in the 1891 census; the marriage register says Jonson; great-grandma’s gravestone says Johnstone; and a distant cousin insists the family was “originally” Johnstone of Annandale. Which spelling is “right”?

The honest answer — and the premise of this guide — is that the question is usually wrong. Until about a century and a half ago, surnames simply did not have one fixed spelling, even for the same person in the same year. What survives in the records is whatever a clerk heard, guessed, misread or standardised on the day. So the modern researcher’s job is not to find “the correct spelling” but to search around the name: building a disciplined list of variants, using sound-based indexes with their eyes open, and falling back on name-free identification when every spelling fails.

As always at Select Surname List, we keep two things separate: the origin story of a surname (a general event, centuries ago, that created one name with many future spellings) and the history of your particular family (a specific line of people whose name may have been recorded a dozen ways across their lifetimes). Spelling sits between them: it is neither the meaning of the name nor proof of kinship — it is mostly paperwork. (See our related guide, Welsh Surnames: Patronymics, Spelling Changes, and History, for the deepest-dive example.)

Key takeaways

• Variation was the norm, not an anomaly. Shakespeare’s own name appears in 25 recorded spellings in his lifetime — and none of his six surviving signatures uses the modern “Shakespeare”.
• Variant ≠ deviant. The Guild of One-Name Studies distinguishes spellings a person or officials demonstrably used (variants) from random clerical and transcription errors (deviants).
• Search by sound, not by letters. Group letters that interchangeable — i/e/y, ee/ea, f/ph/v, s/z, t/th/d — and add the standard variant rules for endings and doubled letters.
• Know your tools. Soundex, Daitch-Mokotoff and Beider-Morse indexes group like-sounding names; wildcards and phonetic search options can find records “exact” searching never will — but each has blind spots.
• When all spellings fail, stop searching for the name. Search for the person: given name, age, address, occupation, associates — and let the family surface by cluster.

1. First, a calibration: what a spelling variation can and cannot tell you

Three cautions before the technique, because they save weeks of wasted searching:

1. A different spelling does not mean a different family. It very often means a different clerk, or the same clerk on a bad day.
2. An identical spelling does not mean the same family — the mirror-image error, just as common.
3. Not every spelling found is equally meaningful. The Guild of One-Name Studies, the standard-setting body for surname research, draws a hard line between a variant — “a name spelling… that the person was known to have used, through signature evidence on wills, marriage bonds etc., or… used by officials on a consistent and persistent basis over a period of years” — and a deviant: “any other spelling recorded, including… only randomly and inconsistently,” plus spellings arising from “enumeration, transcription and indexing errors, both contemporary and modern.” In practice: treat every odd spelling as searchable, but only promote spellings into your family’s documented name when the evidence qualifies.

With that calibration, let’s look at why the spellings moved in the first place.

2. Why surnames were spelled differently

Spelling was fluid — then it froze

Spelling in general was unstandardised for most of recorded history; surnames simply inherited that looseness. The best-documented surname in the English language proves the point: Shakespeare’s. E.K. Chambers found 83 spellings of the name across the centuries; David Kathman, tabulating every reference from William’s christening (1564) to his death (1616), counted 25 different spellings in about 160 documents — led by Shakespeare (71 occurrences), then Shakespere (27), Shakespear (16), Shakspeare (13), Shackspeare (12), Shakspere (8) — down to one-offs such as Shagspere, Shaxspere and Shexpere.

The Bard’s own six surviving signatures — all on legal documents from 1612–16 — each spell the name differently (Willm Shakp, William Shaksper, Wm Shakspe, William Shakspere, Willm Shakspere, By me William Shakspeare), and none uses today’s standard form. Contemporaries were no tidier: R.C. Churchill notes that Sir Walter Raleigh appears in period records as Raleigh, Raliegh, Ralegh, Raghley, Rawley, Rawly, Rawlie, Rawleigh, Raulighe, Raughlie or Rayly, while Thomas Dekker appears as Dekker, Decker, Deckar, Deckers, Dicker, Dickers and Dyckers — and, delightfully, Dickens.

Three patterns from the Shakespeare material matter for your research:

• Print normalised; handwriting didn’t. Printed title pages overwhelmingly spelled the name one way; manuscripts wandered. Your ancestor’s spelling was set by whoever held the pen.
• The metropolis was tidier than the provinces. Provincial documents show the wildest variation; London legal documents cluster on fewer forms.
• A single “official” record can mislead. The spellings in the Stratford parish register — which look so authoritative — are themselves early-17th-century transcripts made by one copyist using his own preferred forms. The lesson generalises: no single record, however official, defines your family’s name.

The great freeze came with universal record-keeping and literacy: printers’ house styles, dictionaries, and above all the machinery of the modern state — civil registration, the census, schools, and the habit of people seeing their own name in print and defending its spelling. In England and Wales, civil registration (1837) and the census schedules largely fixed each branch’s spelling; in the US, naturalization paperwork (which documented name changes only from 1906) did similar work. After that, a spelling usually “sticks” at one memorable document — often the first birth or marriage certificate — and everything earlier becomes fair game for variants.

Clerks wrote by ear

Before that freeze, the record-keeper — parish curate, tax assessor, enumeratior, immigration inspector — typically spelled the name as it sounded to him, filtered through his own dialect and literacy. Two consequences recur everywhere:

• The spelling of a family can change with the record-keeper. The Guild’s advisers note that “the recorded spelling in parish registers changes with the change of incumbent.” One researcher’s Hollier baptisms at St Martin-in-the-Fields ran: Hollier (1630), Holliard (1632, later buried as Hollyer), Hallier (1638), Hollyard (1639), Hollyer (1642), Hollierd (1646) — several clearly different hands’ worth of one name.
• Variation can appear inside a single document. In the 1592 quarto of Arden of Faversham, the character Shakebag’s name — chosen precisely because it parallels “Shakespeare” — appears six different ways across 43 occurrences, sometimes Shakbag and Shakebagge two lines apart. That document-level fickleness shows how interchangeable these spellings were: nobody thought Shak contradicted Shake.

Bilingual layers: translation, not just respelling

Names also changed when they crossed a language boundary — sometimes by translation rather than respelling. In Wales, which we cover in depth in our Welsh surnames guide, clunks like Ddu (“black”) became Dee and Caradog became Craddock because English record-keepers wrote what served English practice. The same forces anglicised Huguenot, Dutch and German names: Schmidt → Smith is respelling-plus-translation; Orlando shading toward Rowland shows translation in the other direction. When your surname “doesn’t look Welsh/Irish/German whatever it’s supposed to be,” ask whether it was translated, not merely mangled.

Migration and transliteration

Names moving between alphabets accumulated changes at every pass: Cyrillic, Greek, Hebrew or Chinese script to Latin is an act of approximation, and each copyist along the route (ticket agent in the village, port clerk, shipping line, arrival inspector, census-taker) had another chance to alter it. JewishGen’s researchers illustrate how one Yiddish-form name can arrive as Zendle, Zendel, Zindel or Zendell — and how even canonical names scatter: Cohen surfaces as Coen, Cowan, Cohan, Cone, Kon, Kohn, Kaan.

“They changed it at Ellis Island” — mostly, no

One of the most persistent family legends deserves a word, because it shapes how people search. The evidence — marshalled by the US immigration service’s own historian Marian L. Smith, the New York Public Library, and Smithsonian — is that Ellis Island inspectors did not write down or alter names. Ship manifests were created abroad, when the ticket was bought; the inspector checked the passenger against the list already in front of him. Inspectors were themselves multilingual (from 1892–1924 roughly a third were foreign-born, averaging three languages each), and there was no form, process or authority for a names-change transaction. The far more common reality: immigrants changed their own names afterwards, often within five years of arrival — a fact so well known that Congress required name-change documentation in naturalizations from 1906. (A near-exception, Frank Woodhull in 1908 — recorded as “Mary Johnson” after a gender-identity detention at the island — only proves the rule, since it ended with the passenger walking out as Frank Woodhull.) For searching, the practical consequences are good news: the arrival manifest usually preserves the embarkation-era spelling, so both ends of the journey — the old-country and the new-country documents — must be searched in their own right.

3. Five kinds of spelling variation (and how to treat each)

4. Build your variant list — a systematic method

Hoping to stumble on the right spelling is not a method. Instead, construct the list deliberately:

1. Say the name aloud — then say it in period dialect. Strong local accents were the engine. Accepted wisdom among surname students is that vowel sounds drift and consonants persist, which tells you where to hunt: vowels and vowel-spellings first.
2. Apply interposable-letter groups. The recurring swaps, in rough order of usefulness:

3. Collapse “fancy” spellings back to sound — and search both. Some surnames are spelled one way and said another, so clerk spellings leap to the pronunciation: Mainwaring → Mannering/Maning; Cholmondeley → Chumley/Cholmley; Featherstonehaugh → Fanshaw/Fanthill; St John → Sinjin/Sentjohn; Marjoribanks → Marchbanks. If your name has a traditional local pronunciation, half your ancestors may be indexed under it.
4. Vary the given names too. Nicknames were formal: Sarah / Sally, Mary / Polly / Molly, Agnes / Nancy / Ann, Hannah / Ann / Nancy, Elizabeth / Betsy / Betty, Edward / Ned / Ted. An entry that “isn’t your ancestor” because it reads Polly is exactly the sort of record exact-name searches lose.
5. Qualify each spelling before you trust it — apply the Guild’s test. A spelling your ancestor signed, or one used consistently by officials for years, is a genuine variant; a stray form in one tax roll is a deviant to be searched, not canonised.
6. Standing advice from one-name researchers: collect the variant fan, but register/study the handful that matter — the Guild recommends starting with about the five commonest true variants and expanding from evidence.

5. Use the sound-alike tools — with open eyes

Modern databases give you three families of help. Learn their mechanics and their blind spots.

Wildcards. Most genealogy platforms let * stand for any number of letters and ? for one — e.g. Holli* retrieves Hollier, Hollyer, Holliard in one pass. Rules vary: JewishGen, for example, restricts wildcards to one search mode and offers brackets to force a first letter ([F]isch finds Fish but not Visz). Always read each site’s search-help page; the single most common beginner error is over-typing everything — name, place, dates — so exact filters bury the mangled record. Search broad first, narrow one field at a time.

Soundex (NARA Soundex). Devised by Robert Russell in 1918 and used for the 1930s WPA indexes of the US censuses, Soundex reduces a name to letter + three digits by sound: vowels A, E, I, O, U, H, W, Y are ignored; B/F/P/V=1, C/G/J/K/Q/S/X/Z=2, D/T=3, L=4, M/N=5, R=6; doubled letters and same-code neighbours count once. Thus Smith and Smyth file together, as do Sherman, Schurman, Shireman, Shurman and Szerman (all S655). Soundex is built into many US record searches (censuses, some immigration indexes), and a calculator will hand you the code. Know its quirks: it ignores leading-sound differences (Kane vs Cohen can hide from each other), so treat it as one tool among several.

Daitch-Mokotoff Soundex. Created by Randy Daitch and Gary Mokotoff of the Jewish Genealogical Society specifically because Russell Soundex fails on Slavic and Yiddish names, D-M uses six digits and handles clustered sounds. Its groupings are aggressive and useful: Auerbach = Ohrbach (097500), Lipshitz = Lippszyc (874400), Lewinsky = Levinski (876450). It can miss things the broader tools catch, so JewishGen researchers recommend running both.

Beider-Morse. The newer phonetic algorithm used as the default on JewishGen, built to trace one name across German, Polish, Russian, Yiddish and more — so the v/w and tz/z shifts of transliterated records don’t defeat it.

The universal caveat: phonetic indexes widen the net and admit false positives — JewishGen’s own help tiers its “fuzzy” options upward precisely because of the trade-off. Phonetic search is a net, not a method: pair it with your hand-built variant list, always check the original page image rather than trusting the transcription, and expect to confirm identifications with age, place and associates rather than the spelling alone.

6. A record-by-record field guide to spelling traps

7. When indexes fail: stop searching for the name

The most liberating skill in genealogy is dropping the surname entirely and identifying the person another way:

• The cluster (neighbourhood) method. People migrated, married and witnessed in clusters. If you can place the family on a map — a street, a parish, a ship’s party — you can find them under any spelling by looking at who surrounds them. Census neighbours persist across decades; confirm a family by its neighbours, not its name.
• The sibling sweep. A common surname vanishes; a sibling with a rare given name does not. Research sideways to the siblings, then walk back to your direct line through shared family evidence.
• Occupation + given name + place + age. In smaller communities, “coal miner Evan, aged 55, Mountain Ash” may identify a man more reliably than any spelling of his surname.
• Browse what isn’t indexed. Many old records (deed books, older registers, some census images) survive but are only browsable page by page — tedious, and sometimes the only way.
• Use distribution data to target the search. Know where a surname concentrates (county/hundred/household-level studies) before you search wide. This is the approach our Welsh guide demonstrates in detail.
• Pre-compiled variant lists. Surname dictionaries and one-name-study registers often publish the known variant sets, saving weeks. Debbie Kennett’s Surnames Handbook is the practical starting point for organising a study this way.
• Keep a variant log. For every spelling you encounter, record the source, date and whether it passes the Guild test (bearer-used / official-consistent). This log becomes your family’s naming history — the honest bridge between “origin of the surname” and “history of your family.”

8. Two worked cases

Twenty-five spellings, one man

Kathman’s tabulation of the Shakespeare record is the single best demonstration that spelling variation is systemic, not a sign of two people. The same living individual was Shakespeare (71×), Shakespere (27×), Shakespear, Shakspeare, Shackspeare, Shakspere, plus such exotica as Shagspere — and in one 1592 printed text we can watch a parallel name flicker six ways over 43 occurrences. His own six signatures differ, none matching today’s spelling, and even the “official” Stratford register entries are a later copyist’s rendering. Meanwhile his contemporaries fared no better (eleven spellings of Raleigh; Dickens recorded for Dekker). Lessons: expect within-document variance; expect the printed form to be tidier than the manuscript form; expect province to be wilder than capital; and never decide “it’s a different family” on the strength of a spelling.

Eleven spellings in one parish

The Guild one-name study of Vayro shows the variant fan in miniature. In Birstall parish alone, the family surfaces in the registers as Varray, Vairey, Vara, Varah, Varay, Varey, Varo, Varrey, Varry, Vary and Verah — every spelling but not, at first, “Vayro” itself. The study’s surname expert, George Redmonds, warned that a V-for-F substitution would be unusual in the West Riding — precisely the kind of check that keeps a variant list honest. Meanwhile in Coverham, the baptisms of William Vayro (1736), Thomas Pharo (1739), Anne Pharoah (1736–39 era), Margaret Varo (1743) and John Vara (1745) resolve to just two families in two townships. Lessons: the fan can be wide enough to hide the family entirely from surname searching; F/Ph/V families must be searched together; and the true unit of search is family + place, with spelling as a secondary key.

Quick-reference checklist

1. Say the name aloud; note traditional pronunciations and their spellings.
2. Build the variant list from the interposable-letter table; include nickname forms of given names.
3. Search exact, wildcard, then phonetic tools (Soundex / Daitch-Mokotoff / Beider-Morse as appropriate).
4. Always check the original page image; log indexer errors where you find them.
5. Search both ends of any migration (departure and arrival records).
6. Use the cluster method and sibling sweep when the surname fails.
7. Qualify each spelling: bearer-used (signature) > consistent official use > one-off (deviant).
8. Never assume variant = different family, or same spelling = same family.
9. Don’t “correct” spellings in your tree — record them with sources; the wobble is evidence.
10. Update the variant log after every new record; it is your family’s naming history.

FAQ

Is a variant spelling proof of a different family? No. Shakespeare carried at least 25 spellings in a lifetime. Spelling follows the clerk, literacy, accent and standardisation — not blood.

How many variants should I track? Hand-search as many as your interposable-letter table generates; treat as family names only the few that pass the Guild test (signed by the bearer, or consistently used by officials). One-name studies typically register about five primary variants.

Which databases search phonetically? US census and immigration indexes commonly offer Soundex; JewishGen offers Daitch-Mokotoff and Beider-Morse as defaults; most big commercial platforms offer a “sounds like” or wildcard option — check each site’s help pages, as wildcard syntax differs.

What if my name was translated rather than respelled? Work from the meaning: a Welsh Ddu family may appear as Dee, a German Schmidt as Smith. Search the meaning’s translations in addition to sound-alikes.

Should I fix the spelling in my family tree to the “proper” form? No — standardising erases exactly the evidence you will need to link records. Record spellings as found, cross-reference them, and let one chosen “headword” spelling simply be a filing label.

Sources and further reading

• David Kathman, “The Spelling and Pronunciation of Shakespeare’s Name” (shakespeareauthorship.com) — tabulated spellings 1564–1616; E.K. Chambers, William Shakespeare: A Study of Facts and Problems (1930), 83 spellings; six signatures summarised under “Spelling of Shakespeare’s name” (Wikipedia).
• R.C. Churchill, Shakespeare and His Betters (1958) — Raleigh and Dekker spelling lists.
• Marian L. Smith (USCIS historian), “American Names / Declaring Independence”; USCIS “Immigrant Name Changes”; New York Public Library blog, “Why Your Family Name Was Not Changed at Ellis Island (and One That Was)” (2013); Smithsonian Magazine, “Did Ellis Island Officials Really Change the Names of Immigrants?” (2016); Joel Weintraub, “The Ellis Island Name Change Myth” (JewishGen, 2015/2017).
• US National Archives, “The Soundex Indexing System” (rules and coding guide) — archives.gov/research/census/soundex.
• JewishGen: Soundex technical guide (Russell and Daitch-Mokotoff, with Auerbach/Ohrbach, Lipshitz/Lippszyc, Lewinsky/Levinski examples); “Search Guide” (Beider-Morse default, wildcards, fuzzy tiers).
• Guild of One-Name Studies: “Variants and deviants” and “Getting started — variants” (variant/deviant definitions; Hollier register examples); Journal of One-Name Studies Vol. 9 No. 9 (2008) for the Vayro study and George Redmonds on V/F interchange.
• Debbie Kennett, The Surnames Handbook (History Press, 2012) — practical guide to surname and variant research.
• P.H. Reaney & R.M. Wilson, A Dictionary of English Surnames (OUP) — the standard etymological reference; T.J. & Prys Morgan, Welsh Surnames (University of Wales Press, 1985); John & Sheila Rowlands, The Surnames of Wales (1996; updated 2013) — see our related guide, Welsh Surnames: Patronymics, Spelling Changes, and History.

For profiles of individual surnames — meanings, formation routes and regional homelands, with their principal recorded variants — see the corresponding entries on Select Surname List.