An epistemological analysis of the use of reputation as evidence

Abstract

Rules 405(a) and 608(a) of the Federal Rules of Evidence allow the use of testimony about a witness’s reputation to support or undermine his or her credibility in trial. This paper analyses the evidential weight of such testimony from the point of view of social epistemology and the theory of social networks. Together they provide the necessary elements to analyse how reputation is understood in this case, and to assess the epistemic foundation of a reputational attribution. The result of the analysis will be that reputational testimony is extremely weak from an epistemological point of view, and that in many cases there are more reliable substitutes that achieve a similar purpose. The obvious fix, in my view, is to eliminate the use of reputation testimony to support or undermine the credibility, honesty, chastity or peacefulness of a witness

Introduction

In many legal cases, the reputation of a person, a social group, an institution, a company or a product is the central issue at stake. The most common scenario involves libel, slander and product disparagement claims, which are direct attacks on a person, an institution or a company’s reputation. The false light and misappropriation torts in the American legal system are another example. The former is the use of false information to present someone in a ‘false light’ that would be highly offensive to any reasonable person; the latter is the use of a person’s name or likeness for commercial purposes without consent. False endorsement claims violate a person’s right of publicity—a celebrity’s commercial interest in the exploitation of his or her name or likeness—and can affect his or her reputation. Similarly, the negative use of a trademark or of a copyrighted work can affect a brand, a firm or an author’s reputation (Heymann, 2011: 1342–1343).

There is, however, an entirely different use of the concept: reputation as evidence. Rule 608(a) of the Federal Rules of Evidence allows the use of testimony about a witness’s reputation to support or undermine the credibility of his or her testimony in trial. Rules 404 and 405 allow the introduction of character evidence in the form of reputational testimony in certain circumstances. In civil trials, it is generally admissible when character is a substantive issue in the case. Examples include negligent hiring, negligent entrustment and child custody cases. In criminal trials, character evidence is admissible if offered by a defendant as circumstantial evidence or if the defendant attacks the reputation of the victim. Once the defendant has ‘opened the door’ to character evidence, the prosecution can also introduce reputational testimony to rebut the defendant.

Given the importance of the concept, both in defamation and evidence law, one would expect to find a well-developed theoretical understanding of reputation in the legal literature. Unfortunately, the opposite is the case. Judges and legal scholars alike have noted the lack of analysis of the concept (Rolph, 2008: 2), and the courts have not attempted to define it:

Rather, in deciding if particular statements are defamatory, they have inquired whether such statements would be likely to elicit certain specific responses from those to whom the statements were published; a statement is held defamatory if it tends to expose the plaintiff to ‘contempt, aversion, disgrace, or induce an evil opinion of him or deprive him of friendly intercourse in society’ (Harvard Law Review, 1956: 877, citations omitted).

The absence of a definition of reputation in the legal literature is somewhat less worrisome in the context of the law of defamation than it is in the use of reputation as evidence. In the former, the fact finder’s main concern is whether the libellous statement would in general have a negative effect on a person’s normal social and business interactions. In most common law countries, the plaintiff does not have to offer evidence of a prior good reputation or evidence to prove specific damage to it. ¹ No precise understanding of what has been impaired is required; only a general understanding of the effects of particular negative claims on any person’s standing in the estimation of other members of society. The situation is very different when reputation is used as evidence. If, for example, a person’s reputation for truthfulness is admitted in trial because of its probative value in supporting his or her credibility, it is necessary to understand how the probative relation works. What methods can be used to establish the probative fact? How reliable are those methods? What is the real probative value of a person’s reputation as a predictor of conduct on a specific occasion? All of these are epistemological questions that require a more thorough understanding of what reputation is, and an assessment of the epistemic foundation of a reputational attribution.

The purpose of this paper is to present an analysis of reputation from the point of view of social epistemology and the theory of social networks, and to examine the consequences of this analysis for the use of reputation as evidence in the American legal system. To anticipate, my conclusion will be that there is no epistemological justification for the use of reputation testimony under rules 405(a), 608(a), and similar provisions. In many cases, however, there are reliable substitutes that provide reputational information that has a clear probative value. Although the paper focuses mostly on the negative task of bringing out the epistemological weaknesses of reputation testimony, I present in some detail several sources of reputational evidence that are not subject to the epistemological problems discussed here.

The plan for the paper is as follows. In the next section I offer a general analytical framework within which the concept of reputation will be discussed. An important element within that framework is the structure of the social group within which a reputation exists. A witness’s location within the network and the strength of his or her ties to the person he or she is testifying about will have important epistemological consequences. In the third section I adopt an epistemological perspective to discuss the kind of evidence that supports a reputational attribution. This initial study will allow us to see the main elements that come into play in the analysis of the concept. Finally, I examine the consequences of this analysis for the evidential use of reputation as stated in the Federal Rules of Evidence. We will see that there is a wide spectrum of possible sources of reputational information, and that the FRE limit the admissible options to the weakest side of the spectrum. Focusing on the strongest types of evidence will enrich the possible sources of reputational information available in legal processes. Although the paper will focus on the use of reputation in evidence law, I will avail myself of the legal analyses, legislation and jurisprudence about reputation in the law of defamation. I do not think they deal with different concepts and the epistemological analysis of reputation as evidence will profit from adding a second perspective.

An analytical framework for reputation ²

The standard form to attribute a reputation to a person, a group, an institution, a company or a product—a reputational object, for short—will be a reputation statement of the following form:

x has a reputation for P in group G at time t.

Most studies of reputation have focused on the reputation of persons (Goffman, 1967; Kreps and Wilson, 1982; Post, 1986; Rolph, 2008), but reputations are also often attributed to companies and products (Fombrun, 1996; Herbig and Milewicz, 1997; Klewes and Wreschniok, 2010), and to social and institutional groups. ³ There are several reasons to think that the reputation involved is one and the same concept. Reputations are mostly dependent on the views of the members of the community or group within which they exist; they do not depend on some property that is human-specific. Furthermore, a reputational object need not be aware of its own reputation, and many evaluative properties of reputable persons are shared by groups and companies. There does not seem to be a compelling reason to divide the analysis according to the subject or object that possesses a reputation.

P is one of the properties for which x has a reputation. A reputational property is (i) any positive or negative evaluative property, (ii) a character trait, (iii) a personal preference or (iv) a behavioural tendency. Examples include: ‘Siemens has a reputation for being a stable company’ (evaluative property), ‘Jane has a reputation for being confident’ (character trait), ‘Julia has a reputation for liking Matisse’ (personal preference), ‘Jim has a reputation for driving aggressively’ (behavioural tendency). In most legal contexts, discussions about reputation are restricted to positive reputational properties. The law of defamation seeks to protect a person’s good name or a corporation’s goodwill. Likewise, many authors have equated reputation with honourableness, worth or dignity, all of which are positive concepts. However, the conditions that give rise to good reputations are mirrored in the negative case, and bad reputations can be predicated of exactly the same entities as good reputations. I will thus speak indifferently of reputation without marking any difference between the positive and the negative kind.

Reputations always have to be relativized to a group G. A single individual’s private opinion does not constitute a reputation. Reputations must be public and have a social existence. The best way to understand the structure of G is via the social network approach often used in sociology and social psychology. ⁴ Social networks have many properties that can be identified and measured, including their size and density, the strength, balance and transitivity of their links, and the presence of homophily, reciprocity, social equality and other properties commonly found in human societies. Ego networks, which take an individual as the focal node of the network, are the natural starting point for the study of reputation (Crossley et al., 2015). This approach will restrict my analysis to the reputation of persons, but most aspects of the analysis can be reasonably extrapolated to the reputation of groups, companies, and even objects.

Most research on ego networks has focused on the group consisting of those individuals known by the subject who occupies the focal node. These kinds of studies are of little help in the study of reputation because they cover only a small fraction of the group of people among whom a person can have a reputation. A person’s reputation can exist among a large group of people that the subject does not know, but who are able to identify him or her by name or facial image. Instead, I will use the term ‘ego network’ to refer to all those individuals who are able to identify the subject. ⁵ An ego network consists of two tiers:

First tier: People directly acquainted with x. They can observe x’s conduct and verbal behavior directly but are not isolated from reputational information that might reach them indirectly.

Second tier: People not directly acquainted with x. They receive and circulate reputational information about x that has been provided either by someone in the first tier or by someone whose source can be traced back to someone in the first tier.

Reputational information often originates in the first tier. A direct witness of a person’s habits or personality can share that information with others, who then pass it on to others situated further afield in the network. But reputational information can originate in the second tier itself, for example, when someone starts a false rumour about the reputational object. The source of information about x for second-tier members is entirely testimonial. Their willingness to accept the testimonial information they receive, either from first- or second-tier members, will depend on the reliability and trustworthiness of the source. I will return to this issue in the next section.

Many people in x’s ego network know each other, forming clusters based on family ties or because they share similar interests, tastes, political views, occupations, and so on. The degree of interconnectedness among members of a network is the network’s density. It is calculated by dividing the number of actual linkages between any two members of the network by the total number of possible linkages (Van der Poel, 1993). In a very dense network—for example, in an isolated community—someone can have a reputation for P in his or her entire ego network. But reputations normally exist in subsets of the ego network. One can expect a subset that includes family members and their acquaintances, another that includes college friends and their acquaintances, another that includes coworkers and their acquaintances, and so on. It is perfectly possible for a person to have a different reputation in each subset. Lars can have a reputation for being compassionate to his family, funny to his friends and lazy to his coworkers. The size and structure of the subsets are determined by the strength of the ties between the members of the network. The strength of a tie is a function of factors such as spending time together, sharing activities, providing emotional support, sharing intimate information and exchanging reciprocal services and resources. A high number of strong ties tends to result in denser subsets of the ego network (Granovetter, 1973). Furthermore, information shared among people with strong ties tends to stay local and circulate narrowly, while information shared with people with whom the speaker has weak ties will likely reach other clusters (Craik, 2009: 12).

Some members of the network, due to their closeness to x or to their privileged position in the network, concentrate a great deal of reputational information about x. I will call these people ‘rich informants’. Having strong ties with them will provide access to plenty of reputational information about x even to people in the second tier. Some rich informants will possess and distribute true information about x, while others will pass along baseless gossip. In both cases, the information that any rich informant possesses will be partial and selective, and subject to all the well-known cognitive biases amply documented in social psychology (Kahneman et al., 1982). Furthermore, a person’s stated perception of another often has evaluative connotations and is seldom passed around in a neutral fashion. ‘The information commonly exchanged about specific persons is an appraisal as well as a description; indeed, these functions tend to be overwhelmingly intertwined. Over 95% of thousands of terms available in the English language to describe persons are evaluative rather than neutral in valence’ (Craik, 2009: 31–32). A speaker’s tone of voice, her gestures, and other nonlinguistic indicators also contribute to mix evaluative and descriptive elements in the information conveyed. It is reasonable to assume that the same is true about the information commonly exchanged about groups, companies and products.

The evidential basis of reputation

A reputation statement is supported by a reputational claim of the form: ‘x is P’. Objective evidence for the reputational claim ‘x is P’ provides the members of G with reasons to believe that x is not only said to be P, but also that x’s reputation for P is warranted. Support is a gradable concept. The degree of epistemic support for x’s reputation will be directly proportional to the degree to which the evidence backs the reputational claims about x. Well-justified reputational claims will generate well-supported reputations, while weakly-justified reputational claims will generate weakly-supported reputations. Reputations will be unsupported in those cases in which there is no evidence ⁶ available to the members of G for the reputational claim ‘x is P’. Evidence for a reputational claim can be of many different kinds. In this section I will examine different types of evidence and the degree of support they provide.

In evaluating the evidence for a reputational claim, it is essential to take into account a family of well-known biases in character evaluation. There is evidence that negative social information about a person, a group or a company tends to have a greater effect on our behaviour and perception than positive social information (Anderson et al., 2011). Our perceptions can be unduly influenced by a single negative trait, an effect commonly known as ‘the devil’s horns’. Mendez (1984) argues that this bias can lead jurors to overestimate the tendency of an accused to act in accordance with prior reproachable behaviour. Furthermore, negative claims tend to be confirmed with less evidence than positive claims and require more evidence to be disconfirmed (Rothbart and Park, 1986). ⁷ In other words, it is much easier to create and maintain a bad reputation. This negativity bias is a psychological cornerstone in the study of reputation, and its epistemological implications will become evident throughout the paper.

A reputational object x can have both a supported and an unsupported reputation for P, as long as that reputation is relative to different subsets of x’s ego network; also, x can have a supported reputation for P and an unsupported reputation for Q within the same subset. It all depends on whether the members of the subset of the network have evidence available to them for the reputational claims. Now it could happen that the members of G have some evidence that x is P, and some evidence that x is not P. It does not follow that x would have a reputation for both P and not P. In that case, the existence of support for rival claims would either fail to give rise to a reputation for either one or benefit the claim with a negative reputational property because of the negativity bias. It all depends on the context, on how ‘negative’ the reputational property is, on how much the members of the group care for that property, etc. In any case, for x to have a reputation for P in G, there must be a dominant reputational claim to that effect circulating within G. Being a dominant reputational claim just means that most people in G are acquainted with it and (almost) no one is acquainted with its negation.

Now, the existence of a group of individuals who are acquainted with a reputational claim about a person, a group, a company, an institution or a product x is not enough to say that x has a reputation for P in G. The following necessary condition must be fulfilled:

Acquaintance condition

Most members of the relevant social group G (i) are able to identify x, (ii) are acquainted with the reputational claim that x is P, and (iii) are not aware of any credible defeaters for the claim that x is P; moreover, (iv) most members of G believe that most other members of G can identify x and are acquainted with the reputational claim that x is P. ⁸

In terms of group structure, the Acquaintance Condition implies that most members of G must also be members of x’s ego network. It also seems to imply that most members of G must be acquainted with each other in order to confirm that other members are also acquainted with the reputational claims about x. But that is not always the case. When x is a celebrity, every member of the celebrity’s ego network, which can include millions of people, assumes that his or her peers know who the celebrity is, even if the celebrity’s name has never come up in conversation. The second-tier members of the celebrity’s ego network simply make the assumption that given the media exposure of that person, it is highly unlikely that other people have not become aware of the celebrity’s name and exploits.

I want to briefly examine the two extremes of the justification continuum for reputational claims, viz. reputations that are well supported and reputations that lack any evidence for them. By examining these two extremes, some important features of the concept will emerge. Consider the following examples of well-supported reputations.

Reputation systems

Peer-to-peer reputation systems are used to create and promote trust among strangers, usually in online environments (Resnik et al., 2000). Reputation systems employ algorithm-based computations of reputation scores for providers, goods and persons based on opinions held by other members of the same domain. A typical example of a reputation system is the algorithm used to establish the reputation of a seller in the eBay community. In most reputation systems, having a good reputation is tantamount to being reliable or trustworthy. Reputation systems are therefore closely related to trust metrics and trust modelling in real or virtual social networks (Cofta, 2007).

Statistical induction

Well-supported reputations can be the result of an inductive inference based on statistical data. If the graduates from Stuyvesant High School in New York City are consistently admitted into the top universities in the country, it is appropriate to say that the school has a good reputation as a feeder school for top universities. Anyone who is aware of the school’s placement record will have a justified belief about the school’s excellence, and the existence of such evidence is sufficient to secure an objective basis for its reputation.

Expert surveys

Noma, in Copenhagen, has the reputation of being one of the best restaurants in the world. The foundation for that claim is the fact that it was awarded first place for three years in a row in The World’s 50 Best Restaurants Awards, and it took second place in the most recent edition (2020). The ranking is based on the votes of nearly 1000 chefs, restaurateurs and critics who have been carefully selected by the organizers, and whose opinions are respected by their peers. Another example of expert-based surveys are the models used to measure corporate reputation, the most prominent of which are Fortune’s Most Admired Companies List (MAC), the Reputation Quotient (RQ), the Corporate Personality Scale and the Emotion, Feeling, Intention Model (SPIRIT) (Money and Hillenbrand, 2006). Fortune’s MAC model, for instance, surveys CEOs and financial analysts, who base their judgment of a company’s reputation on eight characteristics: innovation, financial soundness, employee talent, use of corporate assets, long-term investment value, social responsibility, quality of management and quality of products and services.

Expert testimony

Authority-based reputation need not be the result of a vote by qualified individuals. The testimony of a single individual can be sufficient to boost the reputation of a person or a company. If Rafael Nadal states in an interview that the tennis balls made by company W are the best he has ever used, and he adds that he is not being paid to say so, the company’s reputation will rise immediately. Although it will be based on the opinion of a single individual, it is still a well-supported reputation because there are virtually no individuals in the world who can have a better basis to offer such a judgment. His testimony will be akin to an expert’s testimony in a court of law.

Given the tiered structure of the social group, the strongest evidence to justify a reputational claim will originate in the people who have first-hand access to x, i.e. the people who constitute the first tier of x’s ego network. The clients of the eBay seller, the parents of the children who attended Stuyvesant High School and the critics who dined at Noma are part of the first tier in the seller’s, the school’s and the restaurant’s reputational networks, respectively. In the case of the tennis balls recommended by Nadal, although many tennis players will also have first-tier access to them, the most relevant factor in the tennis ball’s good reputation is Nadal’s privileged position within the network. He is a rich informant about the properties of the balls. The information he possesses is deemed valuable because of his expertise, and any first or second tier member of the group who reads his recommendation will have a good reason to believe the reputational claim.

Unsupported reputation occupies the opposite end of the spectrum of epistemic support. Here we find two cases. The first case occurs when x has a well-supported reputation for P within a social group G, but some people outside of G, call them G*, who lack any evidence for the reputational claims, also attribute that same reputation to x. Sometimes a rumour, which happens to be true, spreads through a population. No one has any idea how it started or whether it is true. But if the rumour is persistent enough it might be able to change the perception of x within G*. In such cases we must say that x has acquired a reputation in G*. Thus, someone or something can have both a supported and an unsupported reputation for P, provided each one is relativized to a different social group.

The second case occurs when there is no evidence available to anyone to verify the reputational property attributed to x, simply because such evidence does not exist. This could happen for many reasons, including gossip, marketing, product placement, a lie, a misunderstanding or an emotional response to an experience with a company, a product or an individual. But if enough people within the relevant social group repeat the claims, x will end up having a reputation even if x in fact lacks the reputational properties attributed to it in the claims. This is perhaps what Wigmore had in mind when he called reputation ‘the secondhand, irresponsible product of multiplied guesses and gossip’ (1904: vol. 7, §1986).

In practice, both cases of unsupported reputation can be treated as one. The source of the information in both cases is entirely testimonial, in the broad sense of the term. One becomes aware of the relevant information through some indirect source such as another person, a newspaper article, a magazine advertisement or a TV commercial. The reputational object x can have an unsupported reputation for P even among a group of first-tier persons who do not have access to the relevant evidence for the claim that x is P. These reputations are unsupported not because the source of the information is testimonial—after all, expert testimony can give rise to well-supported reputations—but because the testimonial information cannot be traced back to a reliable or a competent source.

Most cases of reputational claims fall somewhere between these two extremes. ⁹ The kind of information that justifies reputational claims can take many different forms. A first, obvious source of justification occurs when the evidence is of statistical nature, as in the case of Stuyvesant High School. ¹⁰ But such information is hard or impossible to come by in most cases. A more common source of information is constituted by a witness’s testimony of representative examples in which the property P in question is displayed in x’s behaviour or features. ¹¹ For instance, if P is a restaurant’s quality, specific examples of outstanding flavours and techniques must be used to justify the reputational claim that the restaurant is of high quality. If P is a person’s stinginess, people will want to hear about actual incidents in which the person displayed a stingy behaviour. Unless the reputational property is extremely vague, as in ‘x is a nice person’, a hearer will expect the speaker to provide specific examples of x’s P-behaviours, especially if establishing x’s reputation is important or beneficial to the hearer for practical reasons.

However, even if this type of evidence is available, further information is required to establish a firm basis for x’s reputation. If we were to determine a product or a person’s reputation based solely on representative examples presented by a witness or a customer, her testimony would amount to nothing more than a narration of her personal experience with the testimonial object. To become justificatory evidence for reputation, a person’s testimony must also include information about the reputational object’s status within the network to which it belongs. The source of that information will naturally be other members of x’s ego network. The witness must provide information about the relevant structure of the social network and the strength of the ties that connect her sources to x within that network. A person’s confident testimony about x’s status will indicate that he or she trusts the testimonial evidence provided by her sources, which in turn, must also be based on specific examples of P-behaviours. In other words, most members of the community must believe that someone has a reputational property in a justified manner. The information is hearsay, but without hearsay the evidence presented will not qualify as reputational.

Notice that this latter type of information is what the Acquaintance Condition requires. The existence of a witness who is able to identify x and who can declare that most members of the relevant social group G are able to identify x fulfills the first two conditions. By testifying about x’s status, the witness implicitly attests to the fact that other members are acquainted with the reputational claim that x is P and are not aware of any credible defeaters for the claim that x is P. What makes such a reputation warranted is the trustworthiness of first- and second-hand testimony about specific instances in which x behaves in ways consistent with having the property P.

Reputation under FRE Rules 405(a) and 608(a)

With these elements in place, we can now return to the legal context and explore the legal implications of the foregoing analysis. To study the use of reputation as evidence I will focus on rules 405(a) and 608(a) of the Federal Rules of Evidence. The rules state:

Rule 405. Methods of Proving Character.

(a) By Reputation or Opinion. When evidence of a person’s character or character trait is admissible, it may be proved by testimony about the person’s reputation or by testimony in the form of an opinion. On cross-examination of the character witness, the court may allow an inquiry into relevant specific instances of the person’s conduct.

Rule 608. A Witness

(a) Reputation or Opinion Evidence. A witness’s credibility may be attacked or supported by testimony about the witness’s reputation for having a character for truthfulness or untruthfulness, or by testimony in the form of an opinion about that character. But evidence of truthful character is admissible only after the witness’s character for truthfulness has been attacked.

Notice that the concept of reputation evidence used in the formulation of these two rules is consistent with the analysis of reputation offered in this paper. Consider first the elements of the foundation for reputation testimony for the character trait of truthfulness or untruthfulness under rule 608(a): (i) the witness offering testimony, call her witness #2, is a member of the same community as witness #1, the person whose reputation is being rehabilitated or impeached; (ii) witness #2 has been a member of that community for a substantial period of time; (iii) witness #1 has a reputation for truthfulness or untruthfulness within that community; (iv) witness #2 knows of witness #1’s reputation; and (v) the witness states that reputation. (Imwinkelried, 2020: §5.06[2–3]). In the case of rule 405(a), only the third element changes to accommodate the fact that testimony is not limited to reputation for truthfulness or untruthfulness: (iii) witness #1 has a reputation for a general, moral, law-abiding character; or a specific relevant character trait (Imwinkelried, 2020: §6.02[2–3]). No personal knowledge of the character trait in question is required, although naturally witness #2 must be able to identify witness #1. Testimony can be based on hearsay, third-party information, rumour, gossip, etc. as long as these five conditions have been established. Reputation testimony is an exception to the rule against hearsay, under FRE 803(21).

Opinion testimony works differently. Its foundation is as follows: (i) the witness is acquainted with the person whose reputation is under consideration; (ii) the witness has had sufficient contact to form an opinion about that relevant character trait; (iii) the witness has an opinion of that particular character trait; (iv) the witness states that opinion. Opinion testimony is thus a first-hand account that is limited to members of the first tier of witness #1’s ego network. Although members of the first tier are not isolated from reputational claims that come from other tiers, the foundation for opinion testimony indicates that the report is about the character trait with which the witness has been directly acquainted. Furthermore, a witness can offer opinion testimony without having any knowledge of the person’s reputation. Given this limitation of opinion testimony, it will require a different analysis than the one I offer here for reputation testimony.

Another necessary clarification has to do with the social group relative to which reputation testimony will be provided. The first element of the foundation for both rules states that witness #2 is a member of the same community as witness #1. The meaning of community in this context has been widely debated in the legal literature. In the 18th and 19th centuries, it meant the geographical area in which a person lived. In People v Colantone (1926), the court acknowledged that in large cities like New York a person can be a stranger to his neighbours, but still have a reputation among his coworkers or associates in his particular activities. Fifty years later, in People v Bouton (1980), the Court held that a person’s reputation ‘may grow wherever an individual’s associations are of such quantity and quality as to permit him to be personally observed by a sufficient number of individuals to give reasonable assurance of reliability’. ¹² In the terminology of this paper, the community in question is simply some subset of the ego network of witness #1, some members of which must belong to the first tier.

How big must the subset of the ego network be for someone to have a reputation? I stated above that a person’s private opinion does not a reputation make; this is an idea that can also be found in jurisprudence. ¹³ But how small can the subset get before it ceases to be a reputation? Consider the case of Dun & Bradstreet Inc. v Greenmoss Builders, Inc. (1984). Dun & Bradstreet, a credit rating agency, sent a mistaken report to five subscribers indicating that Greenmoss Builders had filed for bankruptcy; eight days later the subscribers were informed of the error. Although there was no indication that any of the five individual subscribers shared the information with anybody else, the Court ruled that the company’s reputation had been harmed, that is, that by mailing the reports to those five subscribers, Dun & Bradstreet had generated a bad reputation for Greenmoss Builders. Ehrhardt complains that ‘the personal opinion of those five persons would not be sufficient to establish an individual’s reputation because the group holding the belief would be too narrow’ (1986: 871). I disagree. According to my account, Greenmoss Builders did acquire a bad reputation within the group of people linked together by their interest in the company’s financial status. The five individuals were acquainted with the negative reputational claims about the company, and they could safely assume that other individuals subscribed to Dun & Bradstreet’s reports were also so acquainted. The determining factor is not the size of the group but this mutual recognition of acquaintance with the reputational claims. In my view, if there are two individuals who are acquainted with reputational claims about a third one, and they both believe that the other is so acquainted, we can say that the third member of the group has a reputation.

Knowing the minimal size of the social group where a reputation can exist does not address a different question, namely, how many people in the group should the witness know for his testimony to have any probative value? The answer to this question is highly contextual. On the one hand, if a witness knows many people in the person’s ego network, but they are all in the second tier of the network, it is very unlikely that his testimony about the person’s reputation for truthfulness, for example, will allow fact finders to reach any justified conclusion about that person’s actual character for truthfulness. On the other hand, if the character trait is very specific and it is only displayed in front of a very reduced number of people (e.g. a tendency to drive aggressively), reputational testimony based on the witness’ interaction with most members of that small group will be sufficient to reach a strongly justified conclusion about the character trait. Thus, the probative value of reputational testimony depends both on the number of informants and on their degree of acquaintance with the person. These are determined, in turn, by the kind of reputational property under investigation.

With these clarifications in place, we can now begin the epistemological analysis of rules 405(a) and 608(a). There are two central questions we should ask in the study of reputation testimony. The first one regards the content of reputation testimony. What is the fact that is being reported and how is its existence proved? The second one regards the probative value of reputation testimony. Character evidence seeks to prove that someone is a certain way, not just that he or she is regarded in that way. Naturally the latter is evidence for the former, but it is not identical to it. We must therefore inquire in which cases a finding regarding a person’s reputation is a sufficient basis to establish the corresponding character trait. ¹⁴

To answer these questions, I will examine several cases that bring out some of the epistemological elements that I have presented in the previous sections. In particular, I will use the three methods that guarantee well-supported reputations—reputation systems, statistical induction, and appeals to epistemic authority—to inquire whether fact finders could use them to make reputation testimony a reliable basis to infer the character trait in question. The Federal Rules of Evidence do not specify the accepted methods for proving reputation, so we are free to explore the implications of such methods.

Case 1a: Well-supported reputation based on reputation systems

Suppose that under rule 405(a), A testifies about B’s reputation based only on the information compiled using a reputation system. For example, suppose B sales used books on Amazon.com, and A has been buying books from B for many years. A can testify that B has a reputation for being an honest merchant, not based on his own experience, which would be opinion testimony, but based on B’s rating on Amazon.com. The fact being reported here is double natured. It is both B’s reputation and the numerical result of an algorithm. Since B’s rating is public knowledge, the rational course of action will be to use the numerical result itself as evidence. A’s testimony can be replaced by documentary evidence that shows B’s score in the reputation system, and expert testimony about the workings of the rating system. A’s reputation testimony is superfluous.

Case 1b: Well-supported reputation based on statistical induction

The example of objective reputation based on statistical induction introduced previously involved the placement record of the graduates of Stuyvesant High School. In judging the reputation of persons, however, it is more difficult to find a similar source of inductive evidence. The first condition for such evidence to exist is the direct acquaintance of the witness with the person whose reputation is being established, that is, the witness must be a member of the first tier of the person’s ego network. But that is not sufficient. The witness must have a strong tie with the person in order to observe his or her regular behaviour in different circumstances over a long period of time. But in that case, the witness’ testimony will no longer be reputation testimony but opinion testimony!

There is, of course, a different way of understanding how inductive evidence can support reputation testimony. Suppose A testifies about the reputation of B as a faithful husband exclusively on the basis of her interactions with other members of G. In her many conversations about B, no one has ever mentioned or insinuated that B is unfaithful to his wife. On the contrary, people often praise him as a family man and a loving husband. The empirical base for A’s induction is thus a uniform set of reports, all of which confirm the reputational claim. She can use that base to predict with high confidence that the next person she meets in G will confirm the reputational claims she has been hearing all along. But the question is whether fact finders can use that same inductive base to reasonably infer that B is a faithful husband. The answer will depend on whether the witness’s informants are people with strong ties to B in his ego network. As we saw above, the strength of a tie is a function of factors such as spending time together, sharing activities, providing emotional support, sharing intimate information, and exchanging reciprocal services and resources. People with strong ties to B will be able to provide reliable testimony that will enrich the inductive base. But fact finders need to know this in order to make a well-justified inference. So, in addition to merely testifying on B’s reputation, A will also have to describe the relevant structure of the social network and the strength of the ties that connect her sources to B within that network. None of this is provided in rules 405(a) and 608(a). ¹⁵

Case 1c: Well-supported reputation based on epistemic authority

We saw above two different ways in which testimony based on epistemic authority could provide the basis for an objective reputation: expert testimony and expert surveys. Neither of these methods seems germane in the present context. If witness A takes the stand as an expert to testify about B’s reputation, A’s testimony is no longer reputation testimony but expert testimony. For example, if A is a personality psychologist who testifies as a psychologist about some particular personality trait of B, her testimony will not be admissible unless she focuses on her direct examination of B’s behaviour, and not on what people say about B’s behaviour. The second method, expert surveys, leads to the same result. If the witness is allowed to report the findings of a reputational poll or survey conducted among experts, we are still dealing with expert testimony, in this case not in the form of an expert witness but of an expert panel that endorses a majority opinion.

Another possible alternative worth examining is whether the witness could testify about the results of a public opinion poll, which would provide evidence about a person’s reputation. Historically, trial judges were reluctant to accept public opinion polls as evidence because of the rule against hearsay (Becker, 1991; Sherman, 1964), but they have become routinely admitted under FRE 703. ¹⁶ Furthermore, since reputation testimony is excluded from the rule against hearsay under FRE 803(21), it can be argued that public opinion polls are unequivocally admissible for that purpose. Nonetheless, I think that public opinion polls and surveys are insufficient to warrant an inference from the result of the poll to the fact that the person has the reputational property in question. Even if they have been neutrally worded, have large samples and low margins of error, polls fail to meet the demands of the analysis presented in the previous section, according to which most members of the community must believe that someone has a reputational property in a justified manner. A public opinion poll or survey might give high marks to someone as the result of propaganda or advertisement, hardly a solid basis for justified belief.

In sum, none of the three methods described in this paper to justify well-supported reputational claims can be used under rules 405(a) and 608(a) of the Federal Rules of Evidence. Either they fall under a different category of evidence or involve information about the network that is not required by the rules.

Case 2: Unsupported reputation

Recall that a reputation is unsupported when there is no evidence available to anyone in G to support the reputational claims about x. It only requires acquaintance with the claims circulating in the social group, together with the belief that most members of the group are also acquainted with those claims. If a witness takes the stand to declare about the unsupported reputation of someone, the witness will not be able to say that she knows someone who can verify the reputational claims, and in cross-examination she will not be able to explain why the reputational claims should be believed. The most she can say is that she has heard claims about the person, but whatever she has heard is nothing but unacceptable gossip or rumour. ‘Testimony that a certain rumor is being spread throughout a community will not be admitted without additional foundation because it does not establish that the community believes the rumor’ (Ehrhardt, 1986: 872–873).

Reputation testimony thus seems to be in dire straits from an epistemological perspective. Is there a way to inject a modicum of objectivity into reputation testimony under rules 405(a) and 608(a), even if it is a far cry from the reliable methods used to establish well-supported reputations? Reputations are built on testimony, and perhaps by establishing some conditions on the testimonial links among the members of the person’s ego network we might be able to achieve some degree of objectivity. The underlying idea of the courts seem to be that the collective opinion of a person is not gratuitous, and that whatever reputation the person might have is the effect of many testimonial reports with varying degree of reliability that have found their way to the public sphere. The following passage from the majority opinion in Michelson v United States (1948) illustrates this idea:

The witness is, however, allowed to summarize what he has heard in the community, although much of it may have been said by persons less qualified to judge than himself. The evidence which the law permits is not as to the personality of defendant but only as to the shadow his daily life has cast in his neighborhood. This has been well described in a different connection as ‘the slow growth of months and years, the resultant picture of forgotten incidents, passing events, habitual and daily conduct, presumably honest because disinterested, and safer to be trusted because prone to suspect. (…) It is for that reason that such general repute is permitted to be proven. It sums up a multitude of trivial details. It compacts into the brief phrase of a verdict the teaching of many incidents and the conduct of years. It is the average intelligence drawing its conclusion.’ ¹⁷

Presumably, from ‘the shadow his daily life has cast in his neighborhood’ we are to infer the object, the character trait, projecting that shadow. The courts thus seem to use a person’s reputation as the basis for an Inference to the Best Explanation (IBE). According to IBE, we ought to infer the hypothesis that would provide the best explanation of our evidence (Lipton, 2004). Presumably, the best explanation for a person’s reputation for P is that the person has the reputational property P. That property has manifested itself in uncountable occasions and many of these have been registered in writing or in memory and passed along among the members of the community.

The idea of using IBE in the case of reputation testimony faces a major hurdle: What assures us that the witness is faithfully reporting what people think about the person and not just reporting his or her personal views? Reputation testimony has been called ‘opinion in disguise’ ¹⁸ because there is no guarantee that the witness is reporting the views of the community instead of his own views. Due to the uncertainty about the exact content of the witness’ report, it is not unreasonable to consider the idea of adopting public opinion polls or surveys as a more objective way to determine the reputation of a person. I rejected the use of public opinion polls above as a reliable foundation for a well-supported reputation, but the idea here is merely to use them to find out what that reputation is. The proponent of the evidence must show ‘that the survey was conducted in accordance with the accepted standards of survey research, and that the results are being used in a statistically correct manner’ (Ehrhardt, 1986: 881). Thus, when polls and surveys are used in court, ‘the offering party must call expert witnesses to establish the necessary foundation and to interpret the survey results’ (1986: 880). But this means that reputation testimony based on a public opinion poll becomes, once again, dependent on documentary evidence and expert testimony. This is exactly what happened in Case 1a above. This does not mean that public opinion polls cannot be used to determine a person’s reputation in other circumstances, such as the estimation of damages in a defamation case. But it does mean that using them to obtain a precise description of a person’s reputation makes reputational testimony superfluous. In sum, the dilemma we face is either to accept a high degree of uncertainty about the content of the witness’ report, or to accept that the most reliable method to determine the reputation of a person in a community requires expert testimony, which is something not provided in rules 405(a) and 608(a).

Now, independently of the problem just described, there is another obstacle in the use of IBE, or of any other inferential method, to infer a character trait from reputation testimony. I mentioned above the existence of a negativity bias that affects our attribution of reputational properties to people. There is evidence that negative social information about a person tends to have a greater effect on our judgments than positive social information. Furthermore, negative claims tend to be confirmed with less evidence than positive claims and require more evidence to be disconfirmed. This asymmetry naturally introduces a probative imbalance. On the one hand, testimony about a bad reputation will have a stronger tendency to make fact finders conclude that the person has the corresponding negative reputational property, while testimony about a good reputation will have a weaker effect. On the other hand, a witness will be more confident in her testimony about a bad reputation than in one about a good reputation, even if the evidence for either reputational attribution is comparable, that is, even if it is based on conversations with the same number of persons in the social group, who have the same degree of closeness to the defendant. In consequence, a character for truthfulness or for any other positive trait will be more difficult to prove than a character for untruthfulness or for any other negative trait. A rule of evidence that is implicitly biased in this way is a rule of evidence that needs fixing. The obvious fix, in my view, is to eliminate the use of reputation testimony to support or undermine the credibility, honesty, chastity or peacefulness of the witness.

It is true that when character is a substantial issue in civil litigation, the effect of the negativity bias will be negligible, since the issue is not whether the person is or isn’t truthful, peaceful, chaste or honest, but rather whether it can be proved that the defendant was known for having a particular reputational property when he was hired or entrusted with a vehicle, for example. Since these types of cases involve very specific scenarios involving a limited number of people, it is more effective to use specific instances of conduct to present the evidence. In fact, rule 405(b) explicitly states that ‘when a person’s character or character trait is an essential element of a charge, claim, or defense, the character or trait may also be proved by relevant specific instances of the person’s conduct’. If a more reliable source of evidence is available under the Federal Rules of Evidence, there is no reason to prefer the weaker evidence provided by reputational testimony.

In sum, in this section we have seen that the three methods that guarantee well-supported reputations—reputation systems, statistical induction and appeals to epistemic authority—are better sources of reputational information than reputational testimony. Currently there are no evidence rules that incorporate them directly in a legal process as sources of reputational evidence but, as we have seen, they fall under other common categories of evidence such as documentary evidence, opinion testimony and expert testimony.

Conclusion

The result of the foregoing analysis is that there is no epistemological justification for the use of reputation testimony under rules 405(a) and 608(a). The analysis reveals the impossibility of using reputation testimony to establish the existence of a person’s well-supported reputation for a character trait. Furthermore, even if we adopt a lower standard of proof, I have argued that the probative force of reputation is seriously biased if it is used to determine the credibility of a witness, and that it will be insufficient to establish a specific character trait unless it is supplemented with information about specific instances of the person’s conduct or with information about the structure of the social group and the location of the witness’ informants within the network.

I did not discuss the probative value of opinion testimony in this paper, but prima facie from an epistemological point of view it would fare better than reputation testimony in establishing a person’s character traits. Reputation testimony is preferred to opinion testimony by considerations of practical convenience. The following passage is from the majority opinion in People v Van Gaasbeck:

If a witness is to be permitted to testify to the character of an accused person, basing his testimony solely on his own knowledge and observation, he cannot logically be prohibited from stating the particular incidents affecting the defendant, and the particular actions of the defendant which have led him to his favorable conclusion. In most instances it would be utterly impossible for the prosecution to ascertain whether occurrences narrated by the witness as constituting the foundation of his conclusion were or were not true. They might be utterly false, and yet incapable of disproof at the time of trial. Furthermore, even if evidence were accessible to controvert the specific statements of the witness in this respect, its admission would lead to the introduction into the case of innumerable collateral issues which could not be tried out without introducing the utmost complication and confusion into the trial, tending to distract the minds of the jurymen and befog the chief issue in litigation (418).

McCormick also rejects opinion evidence, not because it is inconvenient, but rather because it is unreliable: ‘Opinion testimony concerning a person’s character has traditionally been inadmissible on the basis that it is too unreliable. It will be tainted by the underlying prejudice and bias of the person expressing the opinion’ (Mosteller et al., 2020: section 43). Wigmore defends the opposite point of view:

A perusal of the records of state trials will show how natural, straightforward and useful was this method of asking after belief founded on personal experience and intimacy. Put any one of us on trial for a false charge, and ask him whether he would not rather invoke in his vindication, as Lord KENYON said, ‘the warm, affectionate testimony’ of those few whose long intimacy and trust has made them ready to demonstrate their faith to the jury, than any amount of colorless assertions about reputation. (Wigmore, 1904: vol. 3, §1986)

I agree with Wigmore’s plea for testimony based on personal knowledge, but not for the rhetorical reasons he adduces. From an epistemological point of view, no one is better positioned to offer testimony about a person’s character than the individuals more closely connected to him or her. Since those first-tier individuals within the person’s ego network are likely to be acquainted with each other, their combined personal knowledge of the character trait will serve as a solid basis to establish an objective reputation for that person. In such cases, the psychological studies of social groups based on categories such as emotional and social support networks (Bernard et al., 1990; Burt, 1984) become relevant in order to determine which members of the ego network are best suited to testify. It is necessary, of course, to take into account the practical obstacles mentioned in People v Van Gaasbeck but given the probative weakness of reputation testimony as stated in rules 405(a) and 608(a), it is time to seek a new balance between practicality and the search for truth.