Against Shieber’s Studies on Competence

Much of what we know comes from being told by others. Doctors tell us about our health, our parents tell us about our childhood, and mechanics tell us about our cars.¹ This makes the epistemology of testimony—how we justifiably form beliefs from what others tell us—a central problem for contemporary philosophy. Reductionists believe that testimonial knowledge is grounded in perception, memory, and inductive reasoning; however, anti-reductionists believe testimonial knowledge cannot reduce to other forms of knowledge.² Elizabeth Fricker’s local reductionism argues that justified acceptance of testimony depends on the hearer’s ability to monitor speakers for trustworthiness and competence with respect to what they tell (a proposition p).³ Joseph Shieber cites a series of studies that allegedly show hearers are unreliable monitors of trustworthiness, deceit, and competence; thus, monitoring cannot reliably justify belief in p.⁴ However, Koenig et al. argue that not all instances of accepting what someone says count as testimonial uptake (believing that p once told) in the philosophically relevant sense.⁵ They distinguish telling, where a speaker addresses a hearer and invites them to take their word, from other speech acts like arguments, explanations, and directives.⁶ While Koenig et al.’s collaborative framework differs from Fricker’s reductionism about testimonial warrant, Koenig et al.’s methodological distinctions about what counts as genuine telling provide criteria for identifying whether studies properly test Fricker’s monitoring requirement.⁷ I argue that Shieber’s empirical challenge to Fricker fails on three grounds.⁸ First, Shieber’s studies do not test Fricker’s model case: they use phone orders, videotaped trials, and street interactions rather than face-to-face tellings where monitoring operates with full resources.⁹ Second, these studies place subjects in overhearer roles or measure compliance with directives rather than belief formation by addressees (those being told) in Koenig’s tellings, so they cannot capture the protest and accountability behaviors that signal whether monitoring is functioning.¹⁰ Third, any study claiming to show hearers are unreliable monitors of competence must use a collaborative rather than evidential design, because Fricker’s monitoring is inherently relational as it assesses whether this speaker is competent about this proposition p.¹¹ Shieber’s studies measure general persuasion or compliance, not proposition-specific competence monitoring, and cannot establish his empirical premise.¹²

To evaluate Shieber’s challenge, I first clarify Fricker’s local reductionism with special attention to what “monitoring” for competence amounts to in her view: a defeater-sensitive, automatic, and normatively evaluable assessment.¹³ Second, I reconstruct Shieber’s argument in “Against Credibility,” emphasizing the insights he derives from each of the studies regarding competence.¹⁴ Third, I reconstruct Koenig et al.’s collaborative model and flesh out the distinctions between addressee/overhearer and argument/genuine telling.¹⁵ Fourth, I establish a litmus test for what would constitute a genuine threat to Fricker, and compare each of Shieber’s studies to these standards.¹⁶ Finally, I briefly note what this implies for empirical tests of testimonial monitoring.¹⁷

Fricker’s Local Reductionism about Testimony

In “A Defence of Local Reductionism about Testimony,” Fricker observes that human testimony is “chronically liable to inaccuracy” because people lie, have biases, and speak from ignorance.¹⁸ Fricker argues that responsible belief in testimony requires that hearers assess if speakers are trustworthy and competent about a particular topic before accepting their testimony p.¹⁹ This is a reductionist view because testimonial justification must be grounded in independent evidence of the speakers’ trustworthiness and competence, rather than being automatically conferred from being told p.²⁰

Testimony is an intentional social act in which a speaker presents a proposition p as true while assuming epistemic responsibility for p’s truth.²¹ When someone tells you that p they are also claiming to know p and asking that you believe p on their say-so.²² Thus, telling p is governed by the norm of assertion: “one should not assert that p unless one knows that p.”²³ Telling differs from other communicative acts like directives, demonstrations, or arguments that provide independent reasons; in these cases, the speaker does not present themselves as assuming responsibility for the hearer’s belief in a proposition on their say-so alone.²⁴

To see what is distinct about Fricker’s ‘local’ view, it helps to contrast it with a simpler Humean model.²⁵ Fricker describes a “brutely Humean” (BH) reductionism which treats speakers like free-throw shooters whose reliability is measured by tracking their shooting percentage.²⁶ Under BH reductionism, you would observe many instances where a particular speaker’s utterances “hit” (matched observable facts) or “missed” (were false), then inductively infer their reliability from this track record.²⁷ Unlike basketball, this approach fails for testimony because the facts a speaker testifies to are not independently observable; if they were, there would be no need for testimony.²⁸ Moreover, speakers do not repeat identical utterances the way a player repeats identical free throws; speakers generate novel sentences about novel situations, preventing the kind of repeated correlation BH reductionism requires.²⁹ Most fundamentally, understanding testimony requires recognizing that speakers are intentional agents making knowledge claims for reasons, not mechanical performers whose outputs we simply track for statistical patterns.³⁰

Instead of treating speakers as mere statistical sources, Fricker appeals to a folk-psychological competence assessment.³¹ Fricker calls this Folk-Psychological Inference to Best Explanation (FPIBE), which deploys our understanding of human minds, motivations, and social roles to explain why a speaker is asserting p.³² FPIBE integrates contextual cues, background social knowledge about role-based competences (doctors know medicine, mechanics know cars), understanding of the speaker’s motivations and character, and perceptual evidence from the exchange itself.³³ Fricker “presupposes face-to-face tellings as the model case” because FPIBE demands that the speaker directly addresses a particular hearer, taking responsibility for the hearer’s belief in p, and because face-to-face exchanges provide the perceptual evidence and interactional feedback that FPIBE requires.³⁴

Furthermore, Fricker requires that this monitoring is defeater-sensitive: the hearer must be “always alert” to signs that undermine the speaker’s competence and withholds belief once a defeater is detected.³⁵ Crucially, this monitoring need not involve conscious deliberation about the speaker’s credentials; what matters is that the hearer’s belief-forming process is reliably sensitive to evidence relevant to the speaker’s competence, even if this sensitivity operates automatically below the threshold of conscious awareness.³⁶ For example, while car owners rarely deliberate about a mechanic’s basic knowledge of vehicles, if a mechanic tries to put diesel in an electric vehicle’s charging port, the vehicle’s owner would recognize this as a defeater and seek another mechanic.³⁷ This demonstrates that monitoring operates through automatic perceptual sensitivity to evidence, not through prior conscious reasoning about credentials.³⁸ This is the kind of everyday competence-sensitive monitoring Fricker thinks underwrites testimonial justification.³⁹ Shieber’s argument is that we do not, in fact, monitor competences reliably.⁴⁰

Shieber’s Empirical Challenge to Monitoring

Shieber’s central claim in “Against Credibility” is that many theories of epistemology about testimony incorporate a monitoring requirement, like Fricker’s detailed above.⁴¹ Both Shieber and I agree that Fricker requires consistent monitoring for trustworthiness and competence.⁴² Shieber defines competence as a reliable connection to the truth of proposition p (a mechanic is competent regarding tire durability).⁴³ Shieber believes psychological research strongly suggests we are not reliable monitors of all three.⁴⁴ This tension can be summarized into an inconsistent triad: (1.1) hearers are unreliable monitors of trustworthiness, deceit, and competence; (1.2) justified testimonial belief is acquired through reliably monitoring trustworthiness, deceit, and competence; (1.3) many hearers acquire justified belief in testimony.⁴⁵ After reviewing evidence that hearers are unreliable monitors for trustworthiness and deceit, Shieber argues that if a philosopher wants to keep (1.2) and (1.3), they will treat competence-monitoring as the most promising reconciliation.⁴⁶

Shieber calls this move “modified monitoring,” which drops monitoring for trustworthiness and deception, requiring only monitoring competence instead.⁴⁷ He argues that hearers rarely monitor for competence directly; instead, hearers rely on the mere “trappings of competence” (titles, uniforms, credentials, and other authority markers).⁴⁸ Furthermore, we engage in this process quickly and automatically, through a “click, whirr” shortcut; we track these trappings over the substance of competence they are meant to carry.⁴⁹ To support this claim, Shieber cites Hofling et al. 1966, a study where twenty-one of the twenty-two nurses almost delivered an excessive dosage of an unapproved drug at the command of an unfamiliar doctor over the phone in violation of the hospital’s written order policy.⁵⁰ The same study surveyed student nurses, the majority of whom claimed that they would not have delivered the drug.⁵¹ Shieber then cites Cooper et al. 1996, and claims “when the witness was introduced as having outstanding credentials, the jury found him almost twice as persuasive when presenting his argument incomprehensibly…the jury found the expert more convincing, far more convincing, when they didn’t understand what he was talking about.”⁵² Shieber believes this indicates an automatic preference for outstanding credentials (trappings), rather than the substance of the testimony.⁵³

Shieber also cites Bickman 1974, and states “subjects (hearers) find interlocutors (speakers) more persuasive when they are dressed in a security guard’s uniform than when they are dressed in normal attire—by a difference in compliance rate of 92% to 42%.”⁵⁴ Shieber finally cites Lefkowitz et al. 1955 and states it “demonstrated a man in a business suit and tie was far more persuasive—by more than three-to-one margin in compliance rate—than the same man in a work shirt and pants.”⁵⁵ Shieber then concludes that these studies do not prove that hearers are unreliable monitors of competence because these trappings may correlate with competence.⁵⁶ However, accepting this conclusion complicates modified monitoring as hearers must be able to justify that these trappings reliably indicate expertise, pushing this view towards intellectualism that requires explicit reasoning incompatible with FPIBE’s automatic monitoring.⁵⁷ If we reject the conclusion, modified monitoring transforms into externalist anti-reductionism.⁵⁸

Koenig, Li, and McMyler on Telling and Trust

Koenig, Li, and McMyler’s critique of the “purely evidential model” in developmental psychology echoes Fricker’s rejection of “brutely Humean” reductionism.⁵⁹ Koenig et al. contend that testimonial learning is an “irreducibly social, jointly collaborative epistemic activity, where speaker and hearer coordinate toward true belief through role-structured responsibilities,” not a solitary inference from behavioral data.⁶⁰ Both argue that treating testimony like inductive evidence (e.g. free throws or footprints) fails to capture the interpersonal structure of telling, the distinctive reasons for belief provided by direct address, and the accountability norms that come with taking someone’s word.⁶¹ This shared diagnosis does not mean Koenig et al. endorse Fricker’s local reductionism.⁶² However, their structural account of telling as requiring direct address, interpersonal responsibility, and accountability norms provides the conceptual resources needed to identify when empirical studies genuinely test testimonial uptake as philosophers understand it, versus when they merely test responses to behavioral cues or evidence.⁶³

As Fricker makes a distinction between the model case of telling and the extended form, Koenig et al. distinguish genuine tellings from speech-acts like arguments, demonstrations, and explanations.⁶⁴ The latter provide reasons rooted in being able to arrive at those conclusions independently, and are equally available to overhearers (bystanders)—they are aimed at helping the audience “see for themselves.”⁶⁵ Koenig et al.’s account of genuine tellings is like Fricker’s model case where the addressee is asked to accept the proposition p deferentially, to take the speaker’s word, so it does not aim to place the hearer in a position to make up their own mind.⁶⁶ This leads to Koenig et al.’s central methodological distinction: only addressees, strictly speaking, have been told that p, and only addressees are positioned to “pass the epistemic buck” back to the speaker and hold them accountable for settling whether p.⁶⁷ Evidence from developmental psychology supports this distinction: children show different endorsement and protest patterns when information is addressed at them versus merely overheard, suggesting they recognize direct address as supplying distinctive interpersonal reasons rather than merely providing stronger evidence.⁶⁸

What do the Studies Show?

Koenig et al.’s addressee/overhearer distinction matters here because Fricker’s local reductionism is a view about testimonial warrant: when a hearer is justified in believing a proposition p on a speaker’s say-so.⁶⁹ A study empirically threatens Fricker only if it satisfies three conditions.⁷⁰ First, it must test testimonial uptake in a telling where participants are genuine addressees so that taking the speaker’s word involves the accountability structure that both Koenig et al. and Fricker emphasize.⁷¹ Second, it must show that hearers’ belief formation is not reliably sensitive to evidence or defeaters relevant to Fricker’s trustworthiness condition—that is, to whether “not easily would S assert that P unless she knew that P.”⁷² Third, the manipulation must separate genuine competence and sincerity from mere trappings of competence and demonstrate systematic belief formation driven by the trappings when competence is held fixed or absent, rather than merely eliciting compliance, persuasion, or attitude change.⁷³ If a study places subjects in an overhearer role, supplies independently assessable reasons (as in argument or explanation), or measures deference and compliance without isolating belief in p, then it targets an evidential compliance mechanism that is distinct from monitoring tellings.⁷⁴ The following analyses apply Koenig et al.’s diagnostic questions to each of Shieber’s studies.⁷⁵

Hofling et al. 1966. Shieber treats Hofling et al. as evidence that “trappings of competence” (the caller’s status as a doctor) prompt nurses to deliver excessive doses despite the presence of defeaters.⁷⁶ Twenty-two nurses working alone during short-staffed shifts received phone calls from an unfamiliar doctor ordering them to immediately administer 20mg of “Astroten” to a patient.⁷⁷ The order violated hospital policy because it was not written, the drug was not on the ward’s stock list, and the dosage was twice the maximum daily dose on the bottle.⁷⁸ Twenty-one of the twenty-two nurses prepared to administer the medication before being stopped by observers.⁷⁹ Graduate and student nurses surveyed about the scenario predicted that they would refuse the order.⁸⁰

For Koenig, this is a directive rather than a telling in the epistemological sense.⁸¹ The doctor is not asserting a proposition p for the nurse to believe (e.g. 20mg is the correct dose), but commanding the nurse.⁸² It is not a genuine telling because the exchange does not center on the epistemic uptake of propositional content, despite the direct address towards the nurse.⁸³ Drug dosages are either highly standardized, or involve clinical judgment. Hofling’s design deliberately removes both: the drug is fake (no standard exists), and the phone call prevents clinical deliberation.⁸⁴

For Fricker, the study departs from Fricker’s model case of face-to-face telling in several ways. First, the phone call eliminates contextual monitoring resources such as tone, hesitation, nonverbal cues, and the possibility of real-time challenge. Second, the doctor is unfamiliar, removing background social knowledge about epistemic competences associated with roles that Fricker describes as central to holistic monitoring. Third, the experimental design maximizes stress: the nurse is alone on shift with no doctors or residents in the ward, the drug must be administered urgently, and the situation is engineered to isolate the nurse from her epistemic community.⁸⁵ Crucially, multiple defeaters were present as described earlier. Fricker’s monitoring is supposed to be defeater-sensitive: the nurses should have responded to something being amiss under her view.⁸⁶ However, Rank & Jacobson 1977’s partial replication reveals a critical flaw: when nurses were familiar with the drug (much more common Valium instead of the fake “Astroten”) and allowed to consult peers, 16 of 18 nurses refused the order and engaged in monitoring behaviors such as double-checking the order, contacting their supervisors, consulting their colleagues.⁸⁷ This suggests Hofling et al.’s original design systematically eliminated valid monitoring resources to the degree of flipping the results completely.⁸⁸ From this view, Hofling 1966 primarily measures pressured compliance under hierarchical constraints, not the uptake of testimony about the correctness of a dosage. The outcome is action (preparing the injection) rather than belief formation. More fundamentally, the study measures what happens when an epistemic community’s collaborative monitoring mechanisms are disabled.⁸⁹ The nurses’ compliance reflects the breakdown of institutional safeguards that normally distribute epistemic responsibility in medical settings—closer to the kind of “jointly collaborative epistemic activity” Koenig et al. emphasize.⁹⁰ In summary, Hofling does not test Fricker-style monitoring in genuine tellings for three reasons. First, it involves a directive rather than a telling. Second, it is phone-mediated rather than face-to-face, eliminating the contextual resources described in Fricker’s model case. Third, the design excludes the peer consultation that Rank & Jacobson demonstrate is a critical monitoring resource that Hofling’s methodology appears to have screened out by eliminating cases where nurses contact their colleagues.⁹¹

Cooper et al. 1996. Shieber argues the results of Cooper demonstrate that jurors choose testimony merely because it comes from a prestigious source (trappings) rather than because they understand it (substance).⁹² Cooper et al.’s design places subjects in an overhearer role rather than as addressees in genuine tellings. Mock jurors watched a videotaped civil trial where expert witnesses testified about whether PCBs caused the plaintiff’s cancer. The plaintiff’s expert testimony was manipulated for complexity (think lesion vs. tumor) and credentials (outstanding vs. moderate), while the defense expert always presented complex testimony with outstanding credentials.⁹³ Crucially, the videotape omitted cross-examination, and participants did not deliberate with one another. Jurors could not interrupt, ask clarifying questions, or signal confusion, silencing the accountability measures that are central to both Fricker’s and Koenig et al.’s accounts.⁹⁴

The speech act itself is closer to argument or explanation than to telling in Koenig’s sense. The expert presents scientific evidence and reasoning about causal relations, which the jurors evaluate for themselves, rather than inviting them to accept a proposition p that PCBs caused/did not cause the plaintiff’s cancer.⁹⁵ In an actual courtroom, expert testimony functions similarly as expert witnesses are meant to provide independently assessable reasons, and jurors are positioned as third-party evaluators rather than direct addressees who can “pass the epistemic buck” back to the expert. The videotaped format reinforces the structural distance from the kind of face-to-face telling Fricker describes as her model case.⁹⁶

The study’s findings directly contradict Shieber’s interpretation. Credentials significantly increased persuasion only when testimony was highly complex. When testimony was less complex, “jurors relied primarily on the content of that testimony, and witness credentials had little impact.”⁹⁷ This interaction is evidence against a global competence-monitoring deficit: it demonstrates that jurors shift between content-based evaluation and credential-based cues depending on whether they process the information. When monitoring resources are sufficient, precisely Fricker’s model case, credentials become irrelevant and jurors track content.⁹⁸ The comprehension measure within the study undermines Shieber’s “trappings over substance” narrative. Overall memory for the stated facts was very high (7.27/10) with no differences across the complexity and credential conditions.⁹⁹ The study reveals that even when jurors retain arguments, complexity impedes the evaluation of their quality, prompting reliance on credentials as a coarser indicator when finer-grained assessment is unavailable.¹⁰⁰ Cooper et al. fails to test Fricker-style monitoring because the subjects are overhearers evaluating arguments rather than addressees taking someone’s word, the videotaped format removes collaborative monitoring mechanisms, and most decisively, the study’s own interaction effect shows that if content evaluation is possible, credentials have little impact and jurors rely on substance.¹⁰¹

Bickman 1974. Shieber cites Bickman as evidence that “subjects (hearers) find interlocutors more persuasive when they are dressed in a security guard’s uniform than when they are dressed in normal attire—by a difference in compliance rate of 92% to 42%,” showing that authority markers drive deference “in ways not obviously keyed to real competence.”¹⁰² The epistemic problem here is that Bickman’s field experiments studied obedience to commands, not testimonial belief formation. Experimenters dressed as civilians, milkmen, or guards stopped pedestrians on Brooklyn streets and delivered directives: “Pick up this bag for me,” “Give him a dime,” and “Don’t you know you have to stand on the other side of the pole?”¹⁰³ These are requests for action, not assertions offered for belief. The compliance rate here indicates compliance with an order, not epistemic persuasion. The uniform functions as a cue to the requesters’ legitimacy, not as an indicator of epistemic competence about any proposition p.¹⁰⁴

Bickman’s own analysis confirms this measures social power rather than epistemic monitoring. After informing the pedestrians they were being surveilled, the compliance rates remained identical. This rules out coercive power (you would expect the results to increase under surveillance) and points towards “legitimate power.”¹⁰⁵ Hearers/pedestrians here defer to the uniform’s perceived right to make requests and not whether the requester is competent about an assertion p. Bickman fails to test Fricker-style monitoring because it measures compliance with directives rather than belief formation from tellings. It also operates in fleeting public encounters and any indication of protest or endorsement goes unrecorded.¹⁰⁶

Lefkowitz et al. 1955. Shieber cites Lefkowitz et al. as showing that “a man in a business suit and tie was far more persuasive—by more than three-to-one margin in compliance rate—than the same man in work shirts and pants,” demonstrating that “even these thin status markers in irrelevant contexts substitute for competence-sensitive assessment.”¹⁰⁷ Lefkowitz et al. studied imitation through social modeling at pedestrian crosswalks, not testimonial belief formation. An experimenter dressed either in a business suit or work clothes stood at a signal and crossed or did not cross during the wait portion of the signal.¹⁰⁸ Researchers recorded whether other pedestrians standing nearby violated the signal. If the experimenter wore a suit, 14% of pedestrians violated the signal, in contrast to only 4% when dressed in work clothes. The experimenter dressed in the suit, posited by Shieber as the persuader, never addressed the pedestrians or made any assertion. No subject was told “it is safe to cross” or “the rule does not apply.”¹⁰⁹ There is no telling in either Fricker’s or Koenig’s sense, no direct address, and no propositional content to believe or reject. The experimenter’s behavior provided a social cue about local norms, “businessmen do that here, so it is probably acceptable or low risk.”¹¹⁰ Even if one generously construed the experimenter’s crossing as implicitly asserting “it’s okay to cross now,” pedestrians would still be mere overhearers observing someone else’s decision rather than direct addressees being told that it’s okay to cross now. Baseline compliance with the signal was 99%, so the study essentially measures what weakens compliance to a rule, not what enables belief formation. The study does not test monitoring because it lacks both address and assertion.¹¹¹

Conclusion

Shieber’s empirical challenge does not undermine Fricker’s local reductionism because the empirical studies he cites do not test the kind of testimonial situation that Fricker’s monitoring requirement targets.¹¹² Once monitoring is understood as a defeater-sensitive and automatic assessment of a particular speaker’s trustworthiness and competence regarding a particular proposition p, a study can only threaten Fricker if it captures the proposition-specific, relational task under conditions that supply the relevant monitoring resources.¹¹³

The surveyed cases fall short in three ways: they depart from Fricker’s face-to-face model case, they often place subjects in overhearer roles or measure compliance with directives rather than belief uptake by accountable addressees, and they do not isolate competence-monitoring as distinct from generic persuasion or deference to “trappings.”¹¹⁴ Empirical work that aims to criticize Fricker-style monitoring must distinguish genuine telling from other speech acts, put participants in the addressee position, and measure belief-relevant accountability behaviors rather than mere action or attitude change.¹¹⁵