Implicit Egotism / Name-Letter Effect: The "Dennis The Dentist" Claim Demolished

In 2002, a paper appeared in the Journal of Personality and Social Psychology with one of the more memorable titles in the modern social-psychology literature: “Why Susie sells seashells by the seashore: Implicit egotism and major life decisions.” The lead author was Brett Pelham, then at SUNY Buffalo. The paper made a startling claim. People, Pelham and his coauthors argued, were disproportionately likely to make major life decisions — where they lived, whom they married, what they did for a living — based on subtle, unconscious matches between the names of those choices and their own names. People named Dennis were more likely to become dentists. People named Louis were more likely to move to St. Louis. People named Florence were more likely to live in Florida. People whose first names started with a particular letter were more likely to marry partners whose first names started with the same letter.

The paper was a media sensation. The New York Times covered it. Time covered it. NPR covered it. The finding — soon nicknamed “the Dennis the dentist effect” — became a textbook example in intro psychology courses of how the unconscious mind subtly shapes even the biggest decisions in our lives. Behavioral economics popularizers picked it up. It was repeated in TED talks. It was, by the standard measures, a hit.

Then in 2011, a Wharton professor named Uri Simonsohn — who would later co-found the methodological-critique blog Data Colada and become one of the most influential figures in the replication crisis — published a paper in the same journal. The title was less catchy: “Spurious? Name similarity effects (implicit egotism) in marriage, job, and moving decisions.” The argument was devastating. Simonsohn had spent two years doing detective work on Pelham’s data. What he found was that the entire pattern — Dennis the dentist, Louis in St. Louis, Florence in Florida, the marriage-initial match — was driven not by any unconscious psychological preference but by a confound that Pelham’s original analyses had not controlled for: the birth-year cohort effect. People named Dennis came from a particular generation. People in dentistry came disproportionately from that same generation. Once you controlled for the simple fact that names and professions both vary with the year you were born, the implicit-egotism effect vanished. In some analyses it reversed.

The story of how implicit egotism rose to textbook status on the strength of a confounded analysis, and how a single methodologically careful paper demolished it, is one of the cleanest cautionary tales the replication crisis has produced. For any CEO or strategist evaluating a “subtle unconscious bias shapes consumer decisions” claim — and they are everywhere in modern neuromarketing — this is the case study to know.

What Pelham 2002 Actually Claimed

The Pelham, Mirenberg, and Jones paper (Journal of Personality and Social Psychology, 82(4), 469-487, DOI: 10.1037/0022-3514.82.4.469) presented ten studies. The studies fell into two clusters.

The first cluster, Studies 1-6, looked at residential choices. Using census records, Social Security Administration data, and various directories, the authors compared the observed number of people with a given first name living in a given city against the number that would be expected if names and cities were independent. They reported that people named Louis were 49% more likely to live in St. Louis than would be expected by chance. People named Mildred were disproportionately overrepresented in Milwaukee. People whose names matched any of the largest twelve US cities’ names were systematically overrepresented in those cities. The pattern extended to states (people named Florence in Florida, people named Virginia in Virginia, people named Georgia in Georgia) and to smaller geographic units. The authors framed this as evidence that implicit egotism — defined as an unconscious preference for things associated with the self — was steering people toward residential choices that matched their own identity-relevant features.

The second cluster, Studies 7-10, looked at career choices. Using membership directories from professional associations, the authors compared the observed number of people with given first names in given professions against expected numbers. They reported that people named Dennis or Denise were disproportionately likely to be dentists. People named Lawrence or Laura were disproportionately likely to be lawyers. People named George or Geoffrey were disproportionately likely to be geoscientists. The same logic extended to surnames matching professions (people named Carpenter being carpenters, people named Baker being bakers).

A follow-up Pelham, Mirenberg, and Jones paper that same year extended the framework to marriage decisions, claiming that people were disproportionately likely to marry partners whose first names or surnames shared initials or other letter features with their own.

The theoretical claim was bold. People have an unconscious preference for things that resemble themselves, and that preference is strong enough — even when operating below conscious awareness, even against the backdrop of all the other considerations that go into major life decisions — to systematically shape where we live, what we do, and whom we partner with. The implication for behavioral economics was clear and exciting: if even decisions of this magnitude are subtly distorted by unconscious self-preference, then surely smaller consumer decisions are too.

The Media And Textbook Adoption

The paper landed well. Within weeks of publication it was covered in the New York Times, Time magazine, and on NPR. Malcolm Gladwell referenced it. Behavioral economics writers picked it up as a striking example of unconscious influence on decisions people assume are rational. The finding was vivid, easy to explain, and seemed to point at something deep about how the mind works.

It also became a staple of intro psychology textbooks. The “Dennis the dentist effect” was the kind of finding that lecturers love — short to describe, memorable, with a punchy nickname, illustrating a point about unconscious processes that would otherwise be abstract. By the late 2000s, it was being cited not just in psychology classes but in MBA marketing courses, in popular books about decision-making, and in design and persuasion training materials. The cited evidence was almost always Pelham 2002, sometimes accompanied by the 2005 review in Current Directions in Psychological Science (Pelham, Carvallo, & Jones, 14(2), 106-110).

The finding fit a broader narrative that was dominating social psychology in the late 1990s and early 2000s — the idea that unconscious, automatic processes drove behavior in ways that conscious deliberation barely touched. This was the same era that produced John Bargh’s elderly-priming experiments, money priming, and a whole research program on what came to be called social priming. Implicit egotism fit comfortably into that intellectual neighborhood: yet another demonstration that unconscious associations were shaping behavior at scales we never noticed.

Almost nobody, in those first nine years, pushed back hard on the methodology.

Simonsohn 2011 JPSP — The Cohort-Confound Demonstration

Uri Simonsohn was, in the late 2000s, a young behavioral economist at Wharton with a particular gift for what came to be called forensic statistics. His talent was for looking at published findings and asking the unfashionable question: what else could explain this pattern? Not what is the cleverest psychological story, but what is the most boring confound that has not been ruled out?

In 2011, he published “Spurious? Name similarity effects (implicit egotism) in marriage, job, and moving decisions” in the Journal of Personality and Social Psychology — the same journal that had published Pelham’s original paper nine years earlier (101(1), 1-24, DOI: 10.1037/a0021990). A companion paper, “Spurious also? Name-similarity effects (implicit egotism) in employment decisions,” appeared in Psychological Science (22(8), 1087-1089, DOI: 10.1177/0956797611413937).

Simonsohn’s core argument was this. Names are not randomly distributed across the population. They cluster by birth cohort. The name Dennis was popular in the 1930s through 1950s and has fallen out of favor since. The name Susan peaked in the 1940s and 1950s. The name Jennifer peaked in the 1970s. If you take a snapshot of a profession today and count how many Dennises are in it, you are not just counting Dennises — you are counting people who were born in the years when Dennis was a popular baby name.

But professions also cluster by birth cohort. The age distribution of dentists is not the same as the age distribution of, say, software engineers. Some professions had hiring booms in particular decades. Some professions filter on educational pathways that themselves had cohort-specific enrollment patterns. The number of practicing dentists in any given decade is shaped by when dental schools expanded, when the GI Bill brought a generation into professional education, when licensing standards changed.

Simonsohn’s insight was that if name popularity varies by birth year, and profession entry varies by birth year, then any correlation between name and profession could be entirely explained by the shared birth-year cohort — with no implicit egotism required. Dennis the dentist is not Dennis because he loves the sound of his name. Dennis is a dentist because Dennis was a popular name in the same decade that produced a lot of dentists.

To test this, Simonsohn rebuilt Pelham’s analyses while controlling for birth cohort. He used Social Security Administration historical name data — which records, by birth year, how many babies got each name — and matched it against professional and residential data with age controls. He also reanalyzed extended versions of Pelham’s own data with the cohort confound explicitly modeled.

The results were stark. Once cohort was controlled, the implicit-egotism effects on profession choice shrank toward zero or disappeared entirely. The geographic effects largely disappeared as well — partly through cohort effects, partly through what Simonsohn called “reverse causality” (in some cases, parents in a given city were more likely to give their children names associated with the city, so the kid named Louis didn’t move to St. Louis; he was named Louis because his parents already lived there). The marriage initial-matching effect, in its strongest forms, also failed to survive proper controls for ethnicity, age, and the simple fact that some name pairings are more common than others for reasons that have nothing to do with implicit preference.

Simonsohn was careful in his framing. He did not claim Pelham had committed fraud. He did not claim the original analyses were ill-intentioned. He claimed — and demonstrated — that the analyses were confounded. The data Pelham had used were real. The patterns Pelham had reported were also real. But the causal story Pelham had attached to those patterns — that they reflected an unconscious psychological preference operating on major life decisions — was not supported once the obvious population-statistics confound was controlled for.

How Name-Cohort Effects Produce “Dennis The Dentist”

It is worth pausing on the mechanism because it is not obvious to people who have not worked with population data before.

Imagine you are looking at the membership directory of the American Dental Association in 2002. You count the Dennises. You divide by the expected base rate of Dennises in the general population. You get a ratio. If the ratio is greater than one, Dennises are overrepresented among dentists.

The problem is in the denominator. What is “the expected base rate of Dennises in the general population”? Pelham’s original analyses essentially computed it as: the share of all Americans named Dennis. But the share of all Americans named Dennis is a weighted average across all living birth cohorts — some of which had a lot of Dennises (the 1940s) and some of which had very few (the 1990s).

The dental profession in 2002 was not made up of a representative slice of all living Americans. It was overweighted toward Americans who were in their forties and fifties — old enough to have finished dental school, young enough to still be practicing. That age group was exactly the age group with the highest historical share of Dennises. So the population denominator (the share of Dennises in all Americans) understated the relevant denominator (the share of Dennises in the age group from which dentists are drawn). Divide by too small a denominator, get a ratio that looks like overrepresentation, and you have an implicit-egotism finding.

Simonsohn showed that essentially every striking example in Pelham’s papers had this structure. Louis in St. Louis was an artifact of the fact that the name Louis was older-skewing and so was St. Louis’s demographic. Florence in Florida was an artifact of Florence being an older-skewing name and Florida being an older-skewing state. Dennis the dentist was an artifact of Dennis being a particular-cohort name and dentistry being a particular-cohort profession. The pattern was real. The explanation was a population-statistics confound.

This is the kind of error that is genuinely easy to make. The original Pelham analyses are not obviously broken to someone reading the paper for the first time. They look like sensible base-rate comparisons. The cohort issue jumps out only when you start asking: what else covaries with both the name and the outcome? And it is exactly the kind of question that the social-psychology methodology of the era was not in the habit of asking.

The Specific Reanalyses

Simonsohn’s 2011 JPSP paper documented the cohort-confound problem across all three of Pelham’s major domains.

For profession choice, Simonsohn examined Pelham’s claim that people named after a profession (Dennis-dentist, Lawrence-lawyer, George-geoscientist) were overrepresented in that profession. Controlling for birth-year cohort using Social Security data and age distributions in professional directories, the apparent effects disappeared or shrank to negligible magnitudes. He also pointed out reverse-causality issues: for surname-profession matches like Baker-baker, the surname Baker itself originated centuries ago from ancestors who were bakers, so the modern correlation between the surname and the profession is partly heritable occupational tradition, not implicit egotism in the modern individual.

For residential choice, Simonsohn examined Pelham’s claim that people were overrepresented in cities whose names matched their own. Controlling for cohort (older people are more likely to live in particular states, older names are more common in older people) and for reverse causality (parents in particular cities are more likely to give particular names), the implicit-egotism effects largely vanished. Some residual patterns survived in specific contexts, but the magnitudes were a tiny fraction of what Pelham had originally reported.

For marriage decisions, Simonsohn examined the claim that people disproportionately marry partners whose first names share initials or other features with their own. He showed that proper controls — matching couples on ethnicity, age, religion, and on the simple frequency of particular name pairings in the general population — eliminated most of the apparent effect. He also developed a methodological framework, later expanded in Data Colada post #36, for how to construct proper “active control” comparisons when studying name effects: rather than comparing the rate at which Eric marries Erica against some general population baseline, you have to compare it against the rate at which similar male names (Joseph, Frank, Carl — names with similar wife-name distributions) marry Erica. Once you do that, the apparent Eric-Erica match effect shrinks.

The companion Psychological Science paper applied similar logic to a specific employment-decisions study and reached the same conclusion: the apparent effect was driven by demographic confounds, not implicit egotism.

Pelham’s Response

Pelham did not concede. In the same 2011 issue of JPSP, he and Mauricio Carvallo published “The surprising potency of implicit egotism: A reply to Simonsohn” (101(1), 25-30, DOI: 10.1037/a0023526). The reply defended the original findings, argued that Simonsohn’s controls were in some cases overcontrolling, and suggested that even if some specific examples were artifactual, the broader phenomenon of implicit egotism — well documented in laboratory studies as a preference for one’s own initials in artificial choice tasks — was still real.

Pelham continued the defense in subsequent work, including the 2015 paper “When Tex and Tess Carpenter build houses in Texas: Moderators of implicit egotism” in Self and Identity (14(6), 692-723). That paper presented new analyses using US census data (1880, 1940) and English census data (1911) that controlled for gender, ethnicity, and education, and reported surviving effects across eleven occupations (baker, barber, butcher, and so on). It also proposed theoretical moderators — that implicit egotism should be stronger for high implicit self-esteem and weaker for low implicit self-esteem.

The broader scholarly community was unconvinced. The 2015 Pelham paper used historical census data from cohorts in which surnames were heavily occupational in origin (the name Baker existed because ancestors were bakers) — making the reverse-causality problem even more acute, not less. The proposed moderators (implicit self-esteem as moderator of implicit egotism) belonged to a literature that was itself experiencing replication problems. The defense felt to most observers like a retreat to ever-narrower claims while the core finding — the dramatic, headline-grabbing “Dennis the dentist” version — had been demolished.

Andrew Gelman, the Columbia statistician, summarized the consensus shortly after Simonsohn’s paper appeared: the original finding had been “debunked,” not in the sense of fraud but in the sense that the original analysis had not controlled for an obvious population-statistics confound, and once controlled for, the effect was not there.

The Current Consensus

Here is the honest current state, two decades after Pelham 2002 and fifteen years after Simonsohn 2011.

Implicit egotism as Pelham originally claimed it — a powerful unconscious preference strong enough to systematically shape major life decisions like marriage, career, and residence — is not supported by the evidence. The patterns Pelham documented are real in the descriptive sense (Dennises really do appear at slightly elevated rates in dentistry directories), but the causal interpretation he placed on them does not survive proper controls. The field has, in practice if not always in textbook updates, moved on.

A narrower, weaker version of implicit egotism may still exist. Laboratory studies have repeatedly shown that people, when asked to rate letters of the alphabet, give modestly higher ratings to the letters in their own initials than to other letters — this is the name-letter effect proper, and it is a robust laboratory finding, though it appears to be measuring something like a mere-exposure effect rather than a deep self-preference. Whether this laboratory effect translates into any meaningful real-world behavioral influence is contested. Most contemporary evaluations are skeptical that it does at the magnitudes Pelham claimed.

Intro psychology textbooks have, for the most part, quietly dropped or sharply qualified the Dennis-the-dentist example. The 2005 Current Directions review, once widely cited, has accumulated critical citations alongside the favorable ones. Pelham and his collaborators continue to publish defenses, but the field has shifted toward skepticism. When implicit egotism gets cited in 2026, it tends to be in two contexts: by Pelham himself and his coauthors defending the original program; or, much more often, as a textbook example of how a methodologically careful critique can demolish what looked like a robust finding.

What This Means For “Subtle Unconscious Bias” Marketing Claims

Here is why this case matters for any CEO or strategist evaluating behavioral-science claims in the wild.

The pattern that produced implicit egotism — researcher observes a real correlation in real data, attaches a vivid causal story, the story is repeated until it becomes received wisdom — is the dominant pattern in neuromarketing and consumer-behavior popularization. The structural setup is identical. Some firm or some researcher publishes a finding that consumers respond X% more to such-and-such cue. The finding gets repeated in trade press. Conference talks pick it up. Soon it is a “law” of marketing.

In almost every case where these claims have been examined rigorously, the same Simonsohn-style critique applies: what other variable is the cue correlated with? Is it really the cue that’s driving behavior, or is it something the cue happens to be confounded with — demographics, time of day, channel, season, the kind of customer who is exposed to that cue in the first place?

The implicit-egotism case is particularly clean because the confound (birth cohort) is so visible once pointed out, and because the original claim was so vivid. But the methodology lesson generalizes. When you see a behavioral claim of the form “subtle unconscious cue X shapes major decision Y,” your default question should be Simonsohn’s question: what else correlates with X that also correlates with Y? If the answer is “nothing has been controlled for,” you are looking at correlational data with a causal-sounding story attached, and you should weight it accordingly.

This is particularly important for the modern wave of neuromarketing claims that lean on “the unconscious mind drives the decision.” Most of those claims have not been subjected to anything like the controls Simonsohn applied to implicit egotism. The ones that have, more often than not, do not survive.

What This Case Tells Us About Detective-Work Methodology

The deeper methodological lesson of Simonsohn’s 2011 papers — and one of the reasons he became one of the most consequential figures in the replication crisis — is that there is a difference between two kinds of replication.

The first kind is “did the result replicate” — you run the same study on a new sample and see if you get the same numbers. This is the kind of replication most people think of when they think of the replication crisis: the Many Labs projects, the Reproducibility Project: Psychology, the Bargh elderly-priming failures.

The second kind is what Simonsohn did to Pelham, and what he and his Data Colada colleagues have done repeatedly since: forensic reanalysis of the original data with a question about what causally explains the original pattern. This is harder, slower, and more individual-finding-specific than running replications at scale. But it is often more diagnostically powerful, because it can show not just that a finding fails to replicate but exactly why the original was wrong.

For a CEO evaluating a research-backed claim, both kinds of evidence matter. A failed replication tells you the claim doesn’t hold up empirically. A forensic reanalysis tells you what the actual structure of the original error was — which can save you from being fooled by the next version of the same claim.

Implicit egotism is the canonical example of a finding that was demolished not by failed replication but by detective work. It is the case study that established that some claims need that level of scrutiny before they should be acted on.

Sources

• Pelham, B. W., Mirenberg, M. C., & Jones, J. T. (2002). Why Susie sells seashells by the seashore: Implicit egotism and major life decisions. Journal of Personality and Social Psychology, 82(4), 469-487. DOI: 10.1037/0022-3514.82.4.469.
• Pelham, B. W., Carvallo, M., & Jones, J. T. (2005). Implicit egotism. Current Directions in Psychological Science, 14(2), 106-110. DOI: 10.1111/j.0963-7214.2005.00344.x.
• Simonsohn, U. (2011). Spurious? Name similarity effects (implicit egotism) in marriage, job, and moving decisions. Journal of Personality and Social Psychology, 101(1), 1-24. DOI: 10.1037/a0021990. PDF: urisohn.com.
• Simonsohn, U. (2011). Spurious also? Name-similarity effects (implicit egotism) in employment decisions. Psychological Science, 22(8), 1087-1089. DOI: 10.1177/0956797611413937.
• Pelham, B. W., & Carvallo, M. (2011). The surprising potency of implicit egotism: A reply to Simonsohn. Journal of Personality and Social Psychology, 101(1), 25-30. DOI: 10.1037/a0023526.
• Pelham, B. W., & Carvallo, M. (2015). When Tex and Tess Carpenter build houses in Texas: Moderators of implicit egotism. Self and Identity, 14(6), 692-723. DOI: 10.1080/15298868.2015.1070745.
• Simonsohn, U. (2015). How to study discrimination (or anything) with names; if you must. Data Colada, post #36: https://datacolada.org/36.
• Gelman, A. (2011). Dennis the dentist, debunked? Statistical Modeling, Causal Inference, and Social Science. https://statmodeling.stat.columbia.edu/2011/02/09/dennis_the_dent/.
• Implicit egotism — Wikipedia.

Related

• Replication Crisis Hub — the full series on findings that did and didn’t survive scrutiny
• Bargh Elderly Priming — the parallel social-priming story and the Kahneman “train wreck” letter
• Money Priming — another unconscious-cue claim that mostly didn’t replicate
• Daryl Bem Precognition — the JPSP paper that helped trigger the crisis
• Dunning-Kruger Effect — another textbook finding revisited with better statistics

FAQ

Did Brett Pelham commit fraud?

No, and that needs to be said clearly. There is no evidence whatsoever that Pelham fabricated data or behaved unethically. The Simonsohn critique was a methodological critique — that the original analyses did not control for an obvious confound (birth-year cohort effects) that, once controlled for, eliminates the reported effect. This is a different kind of failure from cases like Diederik Stapel or Marc Hauser, where there was actual fabrication. Implicit egotism is a story about a real researcher making a real methodological error that became consequential because the finding was widely adopted before the error was caught. Pelham has continued to publish in good faith defending and refining the original framework. The scientific community has, on balance, sided with Simonsohn — but the dispute is about methodology and inference, not integrity.

Are there any subtle psychological effects of names that actually do replicate?

The narrow laboratory finding called the name-letter effect — that people, when asked to rate letters of the alphabet, give slightly higher ratings to the letters in their own initials — is a reasonably robust finding in controlled lab settings. It appears to be measuring something like a mere-exposure effect for self-relevant stimuli rather than a deep unconscious preference. Whether this laboratory effect has any meaningful real-world behavioral consequence is a separate question, and the honest answer is that the evidence for real-world consequence is much weaker than the evidence for the lab effect. The Pelham program tried to bridge that gap and overshot dramatically.

What about Data Colada? Why does this matter for the broader replication crisis?

Uri Simonsohn co-founded Data Colada (with Leif Nelson and Joe Simmons) shortly after the implicit-egotism work, in 2013. The blog became one of the most influential venues for the kind of forensic methodological critique that the implicit-egotism case exemplified. Data Colada posts have been load-bearing in exposing data anomalies in several high-profile cases, including Francesca Gino’s fabrication at Harvard. The implicit-egotism critique was, in effect, Simonsohn’s calling-card demonstration of what forensic reanalysis could accomplish — and it helped establish the tradition of post-publication scrutiny that the modern replication-crisis era depends on.

What other “unconscious bias shapes consumer decisions” claims should I be skeptical of?

Most of them, until proven otherwise. The pattern that produced implicit egotism — researcher reports a real correlation, attaches a vivid unconscious-bias story, the story spreads — describes a large fraction of the neuromarketing and consumer-psychology popular literature. Specific claims with a similar track record of crumbling under scrutiny include parts of social priming (mostly didn’t replicate), money priming (mostly didn’t replicate), facial feedback (mostly didn’t replicate), and ego depletion (mostly didn’t replicate). When you see a claim of the form “subtle unconscious cue X causes large behavioral change Y,” your default prior should be that it will not replicate at the magnitudes originally reported, until you see specific evidence of pre-registered, well-powered, properly-controlled replications.

How long did it take for the field to update on implicit egotism?

The Simonsohn paper appeared in 2011, nine years after Pelham 2002. As of 2026 — twenty-four years after the original and fifteen after Simonsohn — implicit egotism still appears in some intro psychology textbooks and is still referenced in some popular books, though usually with qualifications. The field’s working researchers have largely moved on, but the popular and educational literature lags. This is typical of the replication crisis: discredited findings persist in textbooks and popular discourse for a decade or more after the methodological community has stopped taking them seriously. If you are evaluating a research-backed claim and the most recent supporting citation is from before 2010, check what has happened since.

Should I distrust ALL behavioral-science findings now?

No, and that would be the wrong lesson. The right lesson is more discriminating: behavioral-science findings vary enormously in evidentiary quality. Some — like the endowment effect, loss aversion in well-defined contexts, Cialdini’s reciprocity principle for sales conversion — have substantial supporting evidence including replication. Others — like implicit egotism, social priming, and many “unconscious bias shapes major decisions” claims — have substantial evidence against them. The skill, as a strategist, is to learn to ask the diagnostic questions: was this pre-registered? Has it been independently replicated? Has someone like Simonsohn or Data Colada done forensic reanalysis? What is the consensus among methodologically rigorous researchers, not just the originators? If you cannot answer those questions, you should treat the claim as a hypothesis rather than a fact.

What is the bottom line for a CEO evaluating “consumers respond to subtle unconscious cues” claims?

Three rules. First, the default magnitude of subtle unconscious cues on consumer behavior is small to nonexistent — much smaller than popularizations suggest. Second, when an effect is reported, the most likely explanation is some demographic or contextual confound that the original study did not control for, not the cute psychological story the study proposes. Third, the only way to know if a particular cue actually affects your particular customers is to run a properly-controlled experiment on your own audience — and even then, expect smaller effects than the literature suggests, with high variance across contexts. Implicit egotism is the cautionary tale for what happens when you build a strategy on the assumption that the headline number is what’s actually there.