Name–letter effect

Background

Systematic letter-preference research began in 1959 with brand-preference studies by Mecherikoff and Horton. These tried to find the relative appeal of letters for use in package labels. Seven letters were presented in pairwise comparison; one group of subjects judged by sound, another by sight. The letters in a few of the pairs were ranked nearly equally in appeal (e.g., “SK”). In extensions of the study, one by Alluisi and Adams, subjects were asked to rank the English alphabet by the pleasantness of the appearance of capital letters. While there wasn’t a great deal of agreement amongst judges (the coefficients of concordance were low), a strong positive correlation was found between a letter's average rank and how frequently it occurs as initial letter of family names.

Zajonc, a social psychologist, published research in 1968 into preferences between pairs of words (e.g., "on" or "off"): In the overwhelming majority of trials the preferred word was also the most common. He also tested preferences for nonsense words and found that people liked them the more they heard them. Zajonc interpreted these results as evidence that mere repeated exposure to a stimulus is sufficient to enhance its attractiveness.

In around 1977 Belgian experimental psychologist Jozef Nuttin was driving on a highway looking at license plates when he noticed that he preferred plates containing letters from his own name. He wondered if people in general would prefer stimuli that are somehow connected to them; a "mere belongingness" as opposed to Zajonc's mere exposure.

Original research

In his lab at the Katholieke Universiteit Leuven Nuttin designed experiments to test his hypothesis that people place a higher value on letters that feature in their name. It was crucial to the experimental design to rule out other factors, particularly mere exposure. If letters in a name are also letters that occur with higher frequency, then a preference for one's own letters might arise from the mere-exposure effect.

Method

To find an effect which ruled out mere exposure, Nuttin created a yoked control design in which two subjects evaluated the same letters, separately. Some of the letters belonged to one subject’s name, and some of the letters belonged to the other subject’s name, while some were random. In this design any difference in preference between subjects would have to be based on whether the letter occurred in their name. For example, for the fictitious pair Irma Maes and Jef Jacobs the first stimulus was A and U: the last letter in Irma's first name and a letter not in her name. Both subjects had to circle the letter they preferred. The next stimulus was M and D: the penultimate letter from Irma's first name and a letter not part of her name. This was repeated for the remaining letters of her first name, and subsequently the letters of her last name, also in reverse order, and finally the letters of both of Jef's names. The infrequent letters QXYZ were never used as non-name letters. Oral instructions were given to feel a preference as fast as possible, without thinking. If the yoked subjects shared a letter, choosing it did not contribute to the measurement.

In the first trial, 38 Dutch-speaking local elementary school girls circled the preferred letters in two yoked lists of letter pairs. A significant preference for own-name versus partner-name letters was found. The second experiment used 98 local university students, to see if more years of reading made a difference. In addition, other aspects were varied in a full factorial experiment with four between-subject binary factors: 1) either pairs or triads of letters; 2) encircling the preferred letter or crossing out the less preferred one(s); 3) the infrequent letters QXYZ included or excluded; 4) own-name letters listed first or last. All 16 conditions gave a name-letter effect, with a stronger effect when QXYZ were included or with crossing out—the latter mainly due to rejection of partner-name letters. No significant difference was found using family name rather than first name or both names. While the effect was strongest for initials, subsequent data analysis revealed a significant effect even without first and last initial.

Discussion

Nuttin concluded that the experiments proved that, independent of visual, acoustical, aesthetic, semantic, and frequency characteristics, letters belonging to own first and family name are preferred above not-own name letters. He framed the effect in the context of narcissism, an overevaluation of one's own attributes, Gestalt theory, and awareness, as reflected in the choice of title of his 1985 article, "Narcissism beyond Gestalt and awareness: the name letter effect", in which "beyond Gestalt" refers to the fact that the names were not shown, only the letters in isolation, and "beyond awareness" to subjects not realising that their name letters were used. Nuttin claimed the effect he found was the first effect to go beyond Gestalt and awareness.

Second study

In 1987 Nuttin published his second study. The experiments he reported on were done in 1984 and 1985, with the help of Hilde Sas. Because of the far-reaching implications on psychological theories of the name-letter effect, Nuttin found it wise to first test its generality and robustness, before setting off on a research programme aimed at understanding the underlying affective and cognitive processes. He wondered whether the effect could be found in all cultural and linguistic communities, or whether the first study revealed an effect that is due to some unknown idiosyncratic aspect of the Dutch language.

Method

Cross-lingual studies were performed at 13 European universities, using 12 different languages, viz. Dutch, English, Finnish, French, German, Greek (the only one with a non-Roman alphabet), Hungarian, Italian, Norwegian, Polish, Portuguese, and Spanish. Because the original yoked design did not lend itself well to long-distance research and standardization, it was replaced by a simpler, easier to replicate experimental design. Subjects were asked to mark the six capital letters they liked most in a randomized list containing all letters of the local alphabet, again without giving it much thought. Rather than just encircling the six letters of choice, subjects had to mark their first preference with 1, their second with 2, etc. The new method was first applied in Belgium, and when results showed the name-letter effect again, it was copied in the other countries. A total of 2,047 subjects participated, all students.

Discussion

Across languages and letters, the average probability of a letter being chosen as one of the six preferred letters was 0.20 for not-own name letters and 0.30 for own name letters. The strongest effects were observed in the Norwegian and Finnish studies. The effects in the Hungarian, Portuguese, and Italian studies were in the predicted direction but did not reach significance. The effect was also found when only looking at the family names, as well as only first names (with the results of the Italian study reaching significance). When only considering the subject's initials, the name–letter effect emerged as very significant in all 12 studies. Compared to 0.21 for not-initial letters, letters that were part of a subject's initials had a chance of 0.46 to be chosen in the top-six liked letters. However, further analysis revealed that the overall name-letter effect is not simply due to initials: when excluding the initials there still was a name-letter effect across all languages.

Nuttin analyzed the data to see if there was a national-letter effect but failed to find any. Norwegian people did not favor the letter N (for Norge) more than the other nationals did, neither did the Hungarians favor the letter M (for Magyarország) more, etc. This lead Nuttin to conclude that individual ownership has affective consequences that are not observed for collective ownership. The data also allowed for an investigation of the role high-frequency isolated capital letters play. Since cars in Austria and Hungary have a sticker displaying their nationality with a capital that does not match their country's name in the local language (A and H respectively), it provided an opportunity to see if visual prominence is an important factor in the name-letter effect. It did not have any impact on people in those countries liking those letters relatively more.

Similar to the original research, the second study also included a task pertaining to disliking letters. Subjects were asked to select the six letters they liked the least. As before, but now across all countries, merely having a letter in one's own name significantly reduced the chances of disliking it. This did reveal an asymmetry in the letter preference hierarchy. While there was a large consensus within each of the 12 languages as to which letters were least preferred, the most preferred letters did not have much consensus at all. This asymmetry is similar to what Czapiński found when analyzing school-trip companion preferences: much consensus as to who were the three least preferred students, but little in who were the three most preferred. The data for least preferred letters in combination with letter frequency posed serious challenges for the mere exposure theory defined by Zajonc. In each of the languages the least frequent letters were over-represented in the set of least preferred letters, whereas the most preferred letters were rarely the highest frequency letters. Nuttin conjectures that the mere-exposure effects found by Zajonc are in fact mere-low-exposure effects, and that it should be redefined to include relative exposure, not just absolute. The exposure counts in the lab studies done by Zajonc (20 to 25 times) are minuscule in absolute terms compared to the real-life observation counts on any letter. For example, the letter W is the rarest in the French language, yet Francophones will have seen it thousands of times, especially Walloons. Unless it is a name letter, W is universally disliked by French speakers.

Characteristics

In her 2014 meta-analysis of dozens of name-letter effect studies Hoorens called the effect robust. Different aspects of the effect have been investigated:

Cause

Various explanations for the name-letter effect have been explored. Since name letters can not directly be manipulated (one would have to change the subject's name to do so), research into causes can at best be quasi-experimental. Hoorens, Nuttin, Herman, and Pavakanun stated that preferably conclusions regarding causality are drawn from converging results obtained from different experimental designs.

Mere exposure

People may simply like most what they see most. Letters that are more frequent in everyday language also occur more often in people's names. Forer, in 1940, and Alluisi and Adams, in 1962, found a positive correlation between the frequency of occurrence of phonemes and letters and their judged attractiveness. Zajonc extended these studies, using foreign symbols and controlling the number of exposures to the stimuli. This lead him to formulate the mere-exposure hypothesis: the more you see something, the more you like it. However, Nuttin's original study proved that it can't be the cause of the name-letter effect, as letters with equal frequency were evaluated differently by letter owners and non-letter owners. In his 1987 article, Nuttin puts forward more evidence against mere exposure as an explanation for the name-letter effect. Whereas the letter B is of low frequency in French and Portuguese and the letter Q of medium frequency, the B is among the highly liked letters and the Q is among the least liked. Similarly, in Polish and Greek the Y is not a low frequency letter, yet still among the least preferred letters. Karasawa and Kitayama added more evidence: in Japanese culture one rarely uses name initials. Yet they found a strong name-letter effect for initials.

Subjective frequency

Name letters may be noticed more and, consequently, perceived to be more frequent or familiar than non-name letters, causing a subjective familiarity mechanism to enhance their attractiveness. Earlier research into the impact of exposure has shown that even when the objective frequency is identical, differences in subjective familiarity still yield differences in affect. Hoorens and Nuttin tested this by asking subjects to rank-order the entire alphabet twice, once according to their letter preference and once according to their estimated letter frequency. Subjects indeed significantly overestimated the frequency of their name letters. However, there was no significant positive correlation between the overestimation of frequency and the name-letter effect. They also asked subjects to rate their own name. Subjects who liked their name had a stronger name-letter effect than those who didn't like their name, but they didn't overestimate the frequency more. Hoorens and Nuttin concluded that there is no support for the subjective frequency hypothesis.

Evaluative conditioning

Another suggestion is that through repeated visual association of the name letters with the name, the liking of the name is automatically transferred to the letters. Martin and Levey defined evaluative conditioning as a variation of classical conditioning whereby a non-cognitive evaluative response is elicited by evaluatively salient stimuli and transferred to neutral stimuli. Given the observation that our own name stands out among others as quite an attractive stimulus, even, as Mack and Rock found, under conditions of inattention, it could be that name-letter preference is caused by evaluative conditioning. Feys set up a controlled study with Flemish subjects, pairing unfamiliar symbols (Japanese Kanji) with own name and other name. He found that with or without evaluative conditioning did not make any difference on subjects liking the Kanji symbol representing their own name more than the one representing another person's name. He concluded that evaluative conditioning is not the primary cause of the name-letter effect.

Subjective ownership

Nuttin ruled out a conscious response strategy of subjects knowingly choosing the letters from their own name. Despite being given a monetary reward and unlimited time, none of the subjects of his original study were able to find a pattern in the stimulus lists.

Mastery pleasure

The letters first learned by a child, commonly their own name, may be associated with a lasting positive affect. Hoorens and Todorova tested this by looking for a name-letter effect in bilingual subjects where their mother tongue alphabet was Cyrillic and their foreign-language alphabet Roman. Because learning a foreign language at a later age does not typically involve extra attention for name letters, there should be no name-letter effect in the foreign-language alphabet if primacy is the driving factor, only in the first-language alphabet. Results of a study with 100 Bulgarian subjects who at a later age learned English, German, French or Spanish revealed a name-letter effect for the Roman alphabet as well as the Cyrillic. They concluded mastery pleasure is not the principle cause. In a follow-up study Hoorens, Nuttin, Herman and Pavakanun tested mastery pleasure in another way. As mastery pleasure is biggest at the beginning and then diminishes over the years, they tested the strength of the name-letter effect among elementary-school children, in a cross-sectional experimental design involving Flemish and Hungarian second, forth, and sixth graders. Instead of finding the effect to decrease with age, they found an increase, thus proving that mastery pleasure is not the principal cause. They also investigated name-letter effects in bilingual Thai subjects, some of whom had learned the English alphabet at the same time as the Thai alphabet, and others who had learned it later. They found it made no difference on the strength of the name-letter effect when students had learned the second alphabet, thereby ruling out mastery pleasure as a co-determinant.

Probable cause

The effect is thought to arise from unconscious, automatic processes of self-evaluation, with different research groups coming at it from two different angles.

Mere ownership

Nuttin frames the cause in terms of ownership, which has roots in economic psychology. The endowment effect has found that people ascribe more value to things merely because they own them. He puts the reason why the name-letter effect exists down to people automatically liking and valuing anything that is connected to them. As such, the name-letter effect is just one case of a more general mere-ownership effect. If that were true, then various verifiable predictions could be made and tested.

Implicit self-esteem

Starting with Greenwald and Banaji in 1995, another group of researchers have framed the cause in terms of social psychology, specifically the self-esteem theory. This group states that the name–letter effect is caused by links with implicit self-esteem, the implicit positive feelings a person has for themselves. The halo effect of self-esteem spreads into any attributes associated with the self, including the name and its letters; already in 1926 did Syz discover that a person's own name is special compared to others, eliciting a physical responses measurable on the skin. It is thought that when a person recognizes the letters in their name that they experience positive feelings caused by implicit self-esteem. These positive feelings reinforce the behavior of unknowingly selecting the letters in the person's own name, producing the name-letter effect.

Reception

Given how surprising the finding was Nuttin hesitated for 7 years before finally going public with it after all, first at a conference of the European Association of Experimental Social Psychology in 1984, followed by the 1985 and 1987 articles mentioned above. The work was met with widespread skepticism, as he had expected. Loosen, a researcher at Nuttin's own university, called the name–letter effect "so strange that a down-to-earth researcher will spontaneously think of an artefact”. Other researchers did not explicitly say that the effect was spurious, but they doubted its psychological relevance. Over the first five years (1985-1989), Nuttin’s 1985 article was cited only once and the effect was studied at no other university with the exception of Ohio State University, where Johnson replicated the effect using American students.

This all changed in 1995, when Greenwald and Banaji pointed out that Nuttin's work was relevant for indirect measurement of self-esteem, which Nuttin himself had actually suggested already. The original study was cited 14 times in the five years between 1995-1999, 50 times between 2000-2004, 114 times between 2005-2009, and approximately 200 times between 2010 and 2014. The name-letter effect is no longer disputed and Nuttin's work has been called "seminal" by Stieger, Voracek, and Formann in their 2012 meta-analysis of 44 publications on the effect. Their meta-analysis found no trace of publication bias.

Application

The name-letter effect has been exploited in measuring self-esteem. Not all aspects of the self may be accessible through introspection, including self-esteem. Where self-evaluation is possible, it is referred to as explicit self-esteem, where not, as implicit self-esteem. The few studies that have directly investigated the relationship between implicit and explicit self-esteem suggested that the correlation between them is small. This is in line with the standard dual process model, in which explicit self-views are reflected by deliberate and conscious processes, in parallel with implicit self-views being reflected by automatic and spontaneous processes. Measures of implicit self-esteem assess the degree to which the self is cognitively associated with positive versus negative thoughts.

The Letter Preference Task is the second-most popular method to measure implicit self-esteem, surpassed only by the Implicit Association Test. The task has been given various names, including Name Letter Preference Task, Name-Letter-Ratings Measure, and Initial Preference Task. There is no standard method for applying the task. The most used method is a letter rating task, which involves having participants judge all the letters of the alphabet. Even within this method there are variations in instruction (how much you like the letters or how attractive), in scale (5-point, 7-point, 9-point), in the order of the letters (random, alphabetical), and in data collection (paper-and-pencil versus computer-based).

There is no standard for calculating implicit self-esteem. At least 6 algorithms are in use to calculate the size of the name-letter effect. These algorithms are typically applied to initials only, but can be used for all name letters.

In their meta-analysis of the name-letter effect, Stieger, Voracek, and Formann recommend using the I-algorithm, as it exhibits the strongest psychometric properties. In her meta-analysis Hoorens does not recommend a single algorithm. She concludes that little is known about how name-letter preference scores obtained from different algorithms relate to the most important psychometric quality of all, validity.

Stieger, Voracek, and Formann recommend that the task is administered twice, that the effects are calculated separately for first-name initial and last-name initial, that the task is accompanied with the birthday number task, and that the instructions focus on liking rather than attractiveness. They suggest it may be useful to use not just initials but all name letters for measuring implicit self-esteem, something which Hoorens named as her most important recommendation. The Letter Preference Task has been used to measure implicit self-esteem in contexts as diverse as depression, physical health, social acceptance, unrealistic optimism, feedback sensitivity, self-regulation, and defensiveness.

Generalizations

Researchers have looked for evidence of the driving force behind the name-letter effect in related areas of preference, both inside and outside the lab.

In the lab

In controlled studies in the lab, Hodson and Olson tried to find evidence of people liking everyday topics (e.g., foods, animals) that matched their name initials. No evidence was found, neither a between-subjects effect (e.g., Judy did not like jam more than Doug), nor within individuals (e.g., Judy did not like jam more than honey). They did discover a small but reliable effect of initials on brand-name liking within individuals (e.g., Hank did like Honda more than non-matching brands). They speculated that brand names are more likely than the other everyday topics to be used to communicate identity to others. Stieger extended this research by looking at buying preferences for product names. He found that people were disproportionally more likely to buy product names matching their initials. The effect mainly occurred for big brands. No correlation between the name-letter effect and the name-letter-branding effect was found.

Wiebenga and Fennis investigated the underlying affective processes involved in producing a link between the self and the target not through name letters but through the personal pronouns "I" and "my". Both existing and fictive brand names with a generic self-referencing pronoun (e.g., iPhone, mySpace) were evaluated more positively than non-self-referencing brand names (e.g., Xbox). The effect behaved like the name-letter effect: when the self was put under threat it became stronger, and it disappeared for people with negative self-evaluations.

A study by Polman, Pollmann, and Poehlman found that sharing initials with members in a group can increase the quality of group work. Studying undergraduate students they found that groups sharing initials performed better than groups that did not. Groups that had a higher proportion of shared initials exceeded groups with a lower proportion of shared initials.

Controversial studies

Researchers have looked for evidence of the driving force behind the name-letter effect outside the lab. A body of controversial research, starting with Pelham, Mirenberg, and Jones, under the umbrella of implicit egotism has investigated whether people unknowingly make life decisions based on their name letters or name. Sceptics say that to claim that a name or name letters influence life decisions is an extraordinary claim that requires extraordinary evidence.

Uri Simonsohn urged researchers to look hard for confounding variables in field data analysis. A prime example of a confounding variable is baby-name popularity. This has fluctuated significantly over the decades. Simonsohn found that although Walter and Dennis are of near-identical popularity in a large sample of US first names of living people, Walter is a relatively old-fashioned name. Simonsohn suggested that when Pelham et al. found a disproportionally high number of dentists whose name started with the letters "Den" compared to with "Wal", they had overlooked that people named Walter would tend to be old, and more likely to be retired. Using both a different data set (Twitter and Google+ accounts) and a different statistical analysis, Kooti, Magno, and Weber found no evidence of people disproportionally having a job that matched their name initials.

Dyjas, Grasman, Wetzels, van der Maas, and Wagenmakers critized the method Pelham et al. used in their analysis of archives of deaths in 23 "Saint cities" in the U.S., such as St. Louis and St. Paul. Whereas Pelham et al. pooled all data together and concluded that people gravitate towards cities matching their first name, Dyjas et al. used Bayesian hierarchical modelling to show that there are several cities where the opposite happens, people moving away from their name Saint city, and conclude that there is no evidence for an overall effect. A different set of cities containing 30 surnames, such as Jackson and Johnson City, did reveal a disproportionally large number of deceased people with city–surname matches. However, Dyjas et al. did not conclude people gravitate towards cities of their surname, but instead cited Simonsohn's argument that many descendants of founders of these cities may never have moved away, a case of reverse causaility.

The possibility of reverse causality also affected Anseel and Duyck's analysis of a large data set consisting of Belgians' last names and the companies they work for. They concluded that people tend to choose for companies that match their initial. But Simonsohn suspected that, like Walt Disney working for Disney, many people work for companies named after themselves or a family member. When he controlled for reverse causality in a large U.S. data set, he could not see any evidence for people choosing to work for companies matching their initial name letter.

A few controversial studies have linked performance to initials. McCullough and Williams found no evidence of a name-letter effect of the letter 'K' in baseball players striking out (recorded on the scoreboard with a 'K'), despite an earlier study by Nelson and Simmons suggesting there was. Nelson and Simmons also had found students with first name initials C or D get lower, name-letter matching grades than others. Again, McCullough and Williams criticised the statistical analysis used and found no evidence to support such a relation.

In response to Simonsohn's critical analyses of their earlier methods, Pelham and Carvallo published a new study in 2015, describing how they now controlled for gender, ethnicity, and education confounds. In one study they looked at census data and concluded that men disproportionately worked in eleven occupations that matched their surnames, for example, Baker, Carpenter, and Farmer, something the New Scientist has coined nominative determinism. Voracek, Rieder, Stieger, and Swami investigated which way the arrow of causality points when it comes to names influencing choice of occupation. They reported that today's Smiths still tend to have the physical capabilities of their ancestors who were smiths. In the researchers' view a genetic-social hypothesis appeared more viable than the hypothesis of implicit egotism effects.

Undisputed studies

Without controversy is research by Chandler, Griffin, and Sorensen into a link between letter names and donations. They analysed the records of disaster relief donations after seven hurricanes (among others, Katrina, Rita, Mitch). They found that people who shared an initial with the hurricane were overrepresented compared to the baseline of initials of the donors. They concluded that people want to overcome some of the negative feelings associated with the shared name and thus donate. Simonsohn suggested that implicit egotism only applies to cases where people are nearly indifferent between options, and therefore it would not apply to major decisions such as career choices, but would to low-stakes decisions such as choosing a charity.