# ALL-CAPS

Canonical page: https://works.battleoftheforms.com/papers/ssrn-3519630/

[p. 1]
ALL-CAPS
ALL-CAPS
YONATHAN A. ARBEL & ANDREW TOLER*
Alabama Working Paper Series, 3519630
ABSTRACT
A hallmark of consumer contracts is long blocks of capitalized text.
Courts and legislators believe that such “all-caps” clauses improve the
quality of consumer consent and thus they will often require the
capitalization of certain key terms in consumer contracts. Some of the most
important terms in consumer contracts—warranty disclaimers, liability
releases, arbitration clauses, and automatic subscriptions—will be enforced
only because they appeared in all-caps in the contract.
This Article is the first to empirically examine the effectiveness of allcaps with respect to the quality of consumer consent. Using an experimental
methodology, the Article finds that all-caps is significantly harmful to older
readers while failing to show any appreciable improvement over regular
print for others. We collect evidence from standard form agreements used by
America’s largest companies and find that, despite—and perhaps because—
all-caps is ineffective, it is widely used in nearly three-quarters of consumer
contracts. Based on these findings and other evidence reported here, this
Article lays out the dangers and risks of continued reliance on all-caps and
calls for abandoning all-caps.
Draft comments welcome at yarbel@law.ua.edu
Or anonymously through this link.
* Assistant Professor of Law, University of Alabama Culverhouse School of Law;
J.D. Candidate, University of Alabama Culverhouse School of Law. IRB Approval 18-OR408. We would like to thank Oren Bar-Gill, Hillel Bavli, Shawn Bayern, Omri Ben Shahar,
Uri Beonliel, Chris Bradley, Kevin Davis, Shahar Dillbary, Meirav Furth-Matzkin, Robert
Hillman, Dave Hoffman, Nancy Kim, Ben McMichael, Mike Pardo, and Steve Shavell for
helpful comments. We are also thankful to participants in the American Law & Economics
Conference, Conference on Empirical Legal Studies, and Contracts Conference XIV. For
research support, we thank Bret Linley, McGavin Brown, and Victoria Moffa.

[p. 2]
2 ALL-CAPS [DRAFT] [VOL. ---
Table of Contents
INTRODUCTION ........................................................................................................ 3
I.CONSPICUOUS DISCLOSURE AND ASYMMETRIC INFORMATION ........... 11
A. THE PROBLEM OF UNINFORMED CONTRACTING ..................................... 11
II.ALL-CAPS IN ACTION: A STUDY OF INDUSTRY PRACTICES .................... 17
A. METHODOLOGY ..................................................................................... 17
B. FINDINGS ............................................................................................... 18
III.ALL-CAPS AND CONSUMER CONSENT: EXPERIMENTAL ANALYSIS ... 20
A. METHODOLOGY ..................................................................................... 22
B. FINDINGS ............................................................................................... 27
IV.EXPLORING ALTERNATIVE JUSTIFICATIONS AND INTERVENTIONS .. 32
A. ALL-CAPS UNDER TIME PRESSURE ......................................................... 32
1. Methodology ................................................................................. 32
2. Findings ......................................................................................... 34
B. SUBJECTIVE SENSE OF DIFFICULTY & READING SPEEDS ......................... 35
1. Methodology ................................................................................. 35
2. Findings ......................................................................................... 36
C. TAKING THE CON OUT OF CONSPICUOUS ................................................ 39
1. Methodology ................................................................................. 39
2. Findings ......................................................................................... 42
V.THE CASE AGAINST UPPERCASE ................................................................... 44
A. ABOLISHING ALL-CAPS ......................................................................... 45
B. STAIRWAY TO HAVEN ............................................................................ 50
C. THE FUTURE OF DISCLOSURE ................................................................. 52
VI.CONCLUSION ..................................................................................................... 56

[p. 3]
2019] ALL-CAPS [DRAFT] 3
INTRODUCTION
All-caps—blocks of fully capitalized text—is a hallmark of modern
contracts.1 Why this is the case, however, is not well-understood. The
investigation presented here suggests that all-caps is a deeply misguided and
unreflective instance of what Robert Hillman called “contract lore,” a set of
ungrounded beliefs that are passed on through the generations of lawyers.2
One of the deepest problems in contract law is the “no-reading
problem.’3 While consumers are cognizant of certain contractual terms—
such as price and quantity—they are often ignorant of the less salient terms
found in the fine print of their contracts.4 As a result, firms can safely tuck
oppressive terms in the fine print—onerous charges, liability waivers for
wrongful harms, automatically renewing subscription periods, limitations of
representations, arbitration provisions, and damages caps. These practices
1 See e.g., Warranties and Online Sales, AMERICAN BAR ASSOCIATION
https://www.americanbar.org/groups/business_law/migrated/safeselling/warranties/ (Sept.
26, 2016), (noting the scope of the practice)).
2 See generally Robert A. Hillman, Contract Lore, 27 J. CORP. L. 505 (2002).
3 Ian Ayres & Alan Schwartz, The No-Reading Problem in Consumer Contract Law, 66
STAN. L. REV 545 (2014). See also Yanees Bakos, Florencia Marotta-Wurgler & David R.
Trossen, Does Anyone Read the Fine Print? Consumer Attention to Standard Form Contracts,
43 J. LEGAL STUD. 1 (2014) (providing empirical data that virtually no consumers read End
Users License Agreements); Shmuel I. Becher & Esther Unger-Aviram, The Law of Standard
Form Contracts: Misguided Intuitions and Suggestions for Reconstruction, 8 DEPAUL BUS.
& COM. L.J. 199, 206 (2010) (providing empirical data that most consumers are not likely to
read contracts ex ante); Clayton P. Gillette, Rolling Contracts as an Agency Problem, 2004
WIS. L. REV. 679, 680 (2004) (“[C]ommentators agree that buyers, or the vast majority of
them, do not read the terms presented to them by sellers.”); Lewis A. Kornhauser, Comment,
Unconscionability in Standard Forms, 64 CAL. L. REV. 1151, 1163 (1976) (“In general the
consumer will not have read any of the clauses, and most will be written in obscure legal
terms.”). For the formatting of conspicuous disclosures generally, see Mary Beth Beazley,
Hiding in Plain Sight: “Conspicuous Type” Standards in Mandated Communication Statutes,
40 J. LEGIS. 1, 1–2 (2014).
4 See Eyal Zamir, Contract Law and Theory: Three Views of the Cathedral, 81 U. CHI. L.
REV. 2077, 2102–03 (““outside of the law-and-economics community, most people would
quite confidently say . . . that hardly a soul reads standard-form contracts.”).

[p. 4]
4 ALL-CAPS [DRAFT] [VOL. ---
are deeply problematic, as hidden terms pull the consent rug from under the
contracting parties’ feet.5
A common solution to the no-reading problem is to require the
conspicuous display of important terms. If consumers do not read the fine
print, the solution is to make the print less fine.6 Making text conspicuous is
believed to increase the quality of consumer consent by signaling the
importance of the underlying text,7 and by making it more accessible.8 Most
famously, the UCC requires that warranty waivers “must be by a writing and
conspicuous.”9 The UCC is joined by a legion of other statutes, which
incentivize the conspicuous display of information by declining to enforce
key terms that are not conspicuously displayed.10
5 Ayres & Schwartz supra note 3, 549-50 (discussing attempts to address the no-reading
problem).
6 See Richard A. Epstein, Contract, Not Regulation: UCITA and High-Tech Consumers
Meet Their Consumer Protection Critics, in CONSUMER PROTECTION IN THE AGE OF THE
‘INFORMATION ECONOMY’ 205, 227 (Jane K. Winn ed., 2006) (“It seems clear that most
consumers . . . never bother to read these terms anyhow: we . . . adopt a strategy of ‘rational
ignorance’ to economize on the use of our time.”); Alleecia M. McDonald & Lorrie F. Cranor,
The Cost of Reading Privacy Policies, 4 I/S 543, 563–64 (2008) (estimating the time required
to read privacy policies at 244 hours per year per consumer).
7 Bakos, supra note 3, at 2 (noting that the use of fine print “may seem unimportant”).
8 The conspicuousness strategy involves an implicit compromise, as highlighting one term
means that other terms would appear less important in comparison. See Sidney DeLong,
Jacques of All Trades: Derrida, Lacan, and the Commercial Lawyer, J. LEGAL EDUC. 131
(1995) (noting that conspicuousness is a relative quality of the text). See also Regulation Z,
12 C.F.R. § 226.1(b) (2011) (mandating conspicuous disclosure of terms and costs of credit,
at the expense of other contractual terms, in order to promote notice to these aspects of the
transaction).
9 U.C.C. § 2-316 (AM. LAW INST. & UNIF. LAW COMM’N AMENDED 2011).
10 See e.g., CAL. BUS. & PROF. CODE §22577(a)–(b) (West 2004) (A link to privacy policy
must appear “in capital letters equal to or greater in size than the surrounding text.”), FED.
TRADE COMM’N, .COM DISCLOSURES: HOW TO MAKE EFFECTIVE DISCLOSURES IN DIGITAL
ADVERTISING 6 (2013), https://www.ftc.gov/sites/default/files/attachments/press-releases/ftcstaff-revises-online-advertising-disclosure-guidelines/130312dotcomdisclosures.pdf
[hereinafter FTC EFFECTIVE DISCLOSURES] (requiring conspicuous disclosure in
advertisements). See also infra notes 45-46.

[p. 5]
2019] ALL-CAPS [DRAFT] 5
All-caps is a widely endorsed method of making a term conspicuous
and thus rendering it enforceable.11 Courts, legislators, and consumer
agencies take capitalized text to be strong evidence, often dispositive, that
the text was read and understood by the consumer. As a result, courts will
enforce some of the most onerous and demanding terms in consumer
contracts based on the sole fact that this term was written in all-caps.12
Illustration: All-Caps
11 Some statutes outright define conspicuous as “type in boldfaced capital letters”. LA. REV.
STAT. ANN. § 9:1131.2 See also FLA. STAT. ANN. § 718.103 . Sometimes, legislators set
language requirements that employ all-caps. See, e.g., 22 NYCRR 208.6, (“The summons
shall have prominently displayed at the top thereof the words CONSUMER CREDIT
TRANSACTION and the following additional legend or caveat printed in not less than 12point bold upper case type: IMPORTANT!! YOU ARE BEING SUED!! THIS IS A COURT
PAPER--A SUMMONS! DON'T THROW IT AWAY!! TALK TO A LAWYER RIGHT
AWAY!! PART OF YOUR PAY CAN BE TAKEN FROM YOU (GARNISHED). . . .IF
YOU CAN'T PAY FOR YOUR OWN LAWYER, BRING THESE PAPERS TO THIS
COURT RIGHT AWAY. THE CLERK (PERSONAL APPEARANCE) WILL HELP
YOU!!”). For enforcement in the courts. See also Bluewater Trading LLC v. Fountaine Pajot,
S.A., No. 07-61284-CIV, 2008 WL 895705, at 5 (S.D. Fla. Apr. 2, 2008); Brosville Cmty.
Fire Dep't, Inc. v. Navistar, Inc., 4:14–cv–9, 2014 WL 7180791, at 4–5 (W.D. Va. Dec. 16,
2014). Disclaimers have been considered conspicuous where “the excluding language [itself
was] in larger type” or capitalized. Armco, Inc. v. New Horizon Dev. Co. of Va., Inc., 229
Va. 561, 331 S.E.2d 456, 460 (1985) (citing Va. Code § 8.1–201(10)); Young, 1994 WL
506403, at 3 (relying on, albeit not citing, Va. Code § 8.1–201(10)). Hammond–Mitchell, Inc.
v. Constr. Materials Co., CL05000082–00, 2008 WL 8200731, at 5–6 (Va. Cir. Ct. Apr. 28,
2008) (“ConRock used the correct differentiating type-all capitals on the reverse side of the
delivery receipt which was referred to on the front of the ticket[.]”); Rorick v. Hardi N. Am.
Inc., No. 1:14-CV-204, 2016 WL 777575, at 2 (N.D. Ind. Feb. 29, 2016);Lease Acceptance
Corp. v. Adams (2006) 724 N.W.2d 724, at 732 272 Mich.App. 209 (enforcing a forum
selection clause, in part, because it was “printed entirely in conspicuous capital letters”).
12 See, e.g., Bruni v. Didion, 160 Cal. App. 4th 1272, 1293, 73 Cal. Rptr. 3d 395, 413
(2008), as modified (Mar. 24, 2008) (finding that an arbitration clause was surprising because
it was not capitalized).

[p. 6]
6 ALL-CAPS [DRAFT] [VOL. ---
A natural question is whether all-caps is effective—is it truly the case
that capitalization of text improves the quality of consumer consent? Allcaps is, after all, a vestige of the days of the typewriter where other forms of
highlighting text were unavailable.13 Surprisingly, despite the venerable
legacy of this contract lore, this question was never really studied. Instead,
courts and legislators have relied on speculation and intuition.
Admittedly, this question may seem too pedestrian, almost technical;
but as Duncan Kennedy argued, the stakes of “merely technical” questions
in contract law can be very significant.14 Consider then a wrongful death case
where the court will deprive the family compensation only because the
contractual waiver appeared in all-caps.15 If all-caps does not have the effects
attributed to it by courts, this would mean that courts have been erroneously
assuming consent where there was none, enforcing onerous terms in myriad
cases, and depriving consumers of recourse based illusory consent.16 The
terms that need to be conspicuous are those that contracts and legislatures
view as especially important, so enforcing them has particularly acute
consequences for consumer welfare. Worse, if it turns out that all-caps is
effective in hiding meaning, then this would suggest that courts have given
their blessing to one of the most common anti-consumer practices.17
Part I of this Article offers the necessary background regarding the
practice of all-caps. One key finding is that the all-caps practice, despite its
reach and significance, is not based on any evidence. Courts and legislators
adopted this policy because they believe it prevents surprise and improves
13 See Mark Sableman, Typographic Legibility: Delivering Your Message Effectively, 17
SCRIBES J. LEG. WRIT. 9, 9-10 (2017).
14 Duncan Kennedy, The Political Stakes in “Merely Technical” Issues of Contract Law,
19 EUROPEAN REV. PRIVATE L. 7 (2001)
15 See e.g., Enserch Corp. v. Parker, 794 S.W.2d 2 (Tex. 1990) (requiring conspicuous
indemnity language)
16 On the goals of conspicuousness, see infra notes 45-47 and the accompanying text. It is
well understood that actual assent to all terms of the contract may be unwieldy, but many
believe that contract law should demand an affirmative showing of consent to material terms.
See Nancy Kim, Clicking and Cringing, 86 OR. L. REV. 797, 800-05 (2008).
17 See also Beazley, supra note 3, at 2 (arguing that firms intentionally obfuscate
disclaimers); Lauren E. Willis, Performance-Based Consumer Law, 82 U. CHI. L. REV. 1309,
1311 (2015) (arguing that firms hamstring the disclosure project through the framing of
disclosures).

[p. 7]
2019] ALL-CAPS [DRAFT] 7
consent, but this belief is not based on any hard evidence. In fact, the limited
evidence that existed when this practice adopted was mostly negative. In
particular, psychologists started investigating the effects of different
typefaces in the 1930s, and found in a series of studies that it impedes reading
speed.18 Admittedly, these studies are limited; partly because they are dated
and did not explore legal texts. And partly because the focus on reading
speed may have some positive effects, as it may theoretically invite more
careful deliberation. Still, that was the best empirical evidence in existence,
and legal doctrine overlooked it. The doctrine also proved robust to growing
expressions of skepticism of this practice among some practitioners, judges,
officials,19 and a few scholars.20
Part II moves to present evidence on the pervasiveness of all-caps “in
the wild.” To this end, we collected the standard form contracts of 500 of the
most popular consumer companies in the US—companies like Amazon and
Uber—and analyzed them. These forms are the basis of hundreds of millions
of individual contracts between consumers and these large companies. We
use this database to generate the first-ever evidence of the pervasiveness of
long blocks of text in consumer contracts; we find that over three-quarters of
these contracts (77%) contain at least one all-caps clause.
18 Miles A. Tinker & Donald G. Paterson, Influence of Type Form on Speed of Reading, 12
J. APPLIED PSYCHOL. 359 (1928). See also Miles A. Tinker & Donald G. Paterson, The Effect
of Typographical Variations Upon Eye Movement in Reading, 49 J. EDUC. RES. 171, 181
(1955); Miles A. Tinker, Prolonged Reading Tasks in Visual Research, 39 J. APPLIED
PSYCHOL. 444 (1955). Some work has also studied the visibility of capital letters from a
distance, from a distance, see MILES A. TINKER, LEGIBILITY OF PRINT, 33-35, 58-59, but such
an investigation is tangential to our purposes here.
19 See e.g., In re Bassett, 285 F.3d 882, 886 (9th Cir. 2002); OFFICE OF INV. EDUC. &
ASSISTANCE, U.S. SEC. & EXCH. COMM'N, A PLAIN ENGLISH HANDBOOK: HOW TO CREATE
CLEAR SEC DISCLOSURE DOCUMENTS 72 (1998) (proposing that text will not be written in allcaps).
20 See e.g., Beazley, supra note 3, at 2; Ruth Anne Robbins, Painting with Print:
Incorporating Concepts of Typographic and Layout Design into the Text of Legal Writing
Documents, 2 J. ASS'N LEGAL WRITING DIRECTORS 108, 127 (2004).

[p. 8]
8 ALL-CAPS [DRAFT] [VOL. ---
Part III moves to test the effectiveness of these omnipresent all-caps
clauses in lab settings which include approximately 570 participants.21 In our
primary experiment, we test the effect of all-caps on the quality of consent.
If all-caps clauses have any behavioral effect, then respondents should be
able to recall terms better when they are presented in all-caps than when the
same terms are presented in normal print.22 To test this hypothesis, we
presented subjects with a detailed contract with multiple paragraphs, adapted
from a common consumer contract for online music services. In the control
group, the entire contract was written in normal print, which we dub here as
“low-caps.”23 The treatment group saw the same contract, with one
difference: a single paragraph was in all-caps. We then asked subjects about
their obligations under the contract and evaluated the accuracy of their
responses.
The evidence shows that all-caps fails to improve consumer consent in
any appreciable manner.24 Indeed, we find statistically significant evidence
that all-caps strongly undermines the quality of consent for older readers.
For illustration, respondents over 55 were 29 percentage points more likely
to misunderstand their obligations when the paragraph was capitalized than
their age peers who read the paragraph in low-caps. These findings suggest
that all-caps may be harmful to older readers and likely fails to improve
consent for all other readers.
We then conduct several exploratory studies in Part IV. We find some
evidence that all-caps is not helpful even under time pressure; that consumers
21 Overall, for all of our studies we recruited almost 1,000 respondents; our sample size
follows the standard in similar studies. Cf., Meirav Furth-Matzkin and Roseanna
Sommers, Consumer Psychology and the Problem of Fine Print Fraud, (Forthcoming, STAN.
L. REV.) (N=300 in largest study and N=100 in smallest); Tess Wilkinson-Ryan, Do
Liquidated Damages Encourage Breach?, 108 MICH. L. REV. 633 (2010) (N=100);
Wilkinson-Ryan, infra note 83, (N=208).
22 We also consider, and reject, the possibility that all-caps is a signal of worse contract
quality.
23 We use the term low-caps to highlight that we are using standard English grammatical
rules which include some capitalization; e.g., in names and the beginning of sentences. The
appendix provides the different contracts presented to the parties.
24 As will be explained, this conclusion is not based on failure to reject the null hypothesis,
rather, on a non-inferiority test of statistical significance. See infra notes 99-101.

[p. 9]
2019] ALL-CAPS [DRAFT] 9
consider all-caps more difficult to read; and that all-caps may take longer to
read. The potentially negative effect on reading times is consistent with
earlier work in psychology that found that reading capitalized blocks of text
takes roughly 13% longer than non-capitalized text.25
We also tested whether it is possible to improve consumer consent using
alternative means. To this end, we tested the effects of three forms of
highlighting text relative to low-caps. We found strong evidence that the
highlighting of a single line of text using boldface has a considerable positive
effect on outcomes. We interpret this finding as suggesting that some forms
of disclosure can be highly effective if they are properly designed. The
proper design, however, requires close consideration and further
experimentation is necessary.
In interpreting these findings and considering their policy implications,
a few caveats are important. First, we do not find—nor do we argue—that
capitalization is always ineffective. We readily admit that a sufficiently
motivated firm or actor might be able to find a combination of capitalization
and formatting that would be effective.26 Our findings and conclusions
should be interpreted as suggesting that standard usage of blocks of all-caps
text is ineffective and may, indeed, be harmful.
Second, lab experiments are subject to some known limitations. To
minimize these concerns, we took special steps to ensure that we only
recruited subjects from the US and that subjects were actually engaging with
our experiments. To that end, we used a special service that collects the
‘digital fingerprints’ of participants and uses geolocation; we implemented a
number of attention checks; and collected a sample that, with a few
differences, represents the general US population.27 Still, external validity is
always a concern, and it should be emphasized that we are not proposing
here any specific intervention. We seek to discover whether all-caps has its
25 See Miles A. Tinker & Donald G. Paterson, Influence of Type Form on Speed of Reading,
12 J. APPLIED PSYCHOL. 359 (1928).
26 While capitalization is rare in the marketing context, a point we emphasis throughout,
one sometimes finds capitalization in the context of brand logos, such as Pepsi’s. See Tony
Stark, Pepsi Logo, LOGASTER (Dec. 16, 2011), https://www.logaster.com/blog/pepsi-logo//.
27 On MTurk, its benefits, and its limitations, see infra notes 82-83 and 95-98 and
accompanying text.

[p. 10]
10 ALL-CAPS [DRAFT] [VOL. ---
advertised effects, despite the lack of any previous supporting evidence.
Even if lab experiments are limited, however, it is important to remember
the claim that all-caps supporters endorse. They implicitly claim that all-caps
has such strong behavioral effects that it would be justified to disclaim
liability for a crippling accident based on capitalization. Strong claims
require strong evidence; the limits of the lab notwithstanding.
Third, as we test recall, reading speed, and subjective feeling of
difficulty, we do not measure other potential justifications for all-caps.28
Fuller famously argued that formal requirements could be helpful in
providing evidentiary, cautionary, and channeling functions,29 and one might
seek to justify all-caps on the basis of such and other non-behavioral
effects.30 Now, these reasons were never carefully articulated, so it is
uncertain that these reasons are coherent or persuasive. It is not even clear
how one might test these presumed effects and if so, in what direction they
might work. But most significantly, there is a strong normative case against
non-behavioral justifications in this context. All-caps is used to show
meaningful consent to especially onerous terms that would not be enforced
but-for the use of all-caps. If one wants to enforce a disclaimer that prevents
the victim of a medical accident coverage only because the term appeared in
all-caps, this reason must be especially compelling. We are hard-pressed to
find such a compelling reason that is divorced from any behavioral effect.
The results of this study, explored in Part V, carry implications for both
current legal policies and the future of disclosure. In terms of current
policies, we believe that there is enough evidence to abandon the reliance on
all-caps. We base our recommendation in part on the force of the positive
28 Consistent with these metrics, the FTC, for example, emphasizes that the goal of
conspicuous disclosure in online advertising is consumer behavior, not formal notice. FTC
EFFECTIVE DISCLOSURES, supra note 10, at 6 (“Whether a disclosure [is clear and
conspicuous] is measured by its performance–that is, how consumers actually perceive and
understand the disclosure within the context of the entire ad”). The UCC emphasizes the
prevention of surprise to the consumer and requires special clear language to be used. UCC
§ 2-316, cmt 1.
29 Lon L. Fuller, Consideration and Form, 41 COLUM. L. REV. 799, 800-801 (1941)
30 In the UCC, context, courts have taken a more formalistic approach. Stephen E.
Friedman, Text and Circumstance: Warranty Disclaimers in A World of Rolling Contracts,
46 ARIZ. L. REV. 677, 688 (2004)

[p. 11]
2019] ALL-CAPS [DRAFT] 11
evidence presented here, which shows that all-caps is harmful to older
readers while not appreciably improving outcomes over normal print. We
also base our recommendation on the negative evidence we uncover in our
analysis of the case law—showing that there was never any contrary
evidence for this longstanding legal practice.31 Most damning is the
observation that in designing marketing materials, where firms have an
interest in persuading consumers, the use of all-caps is effectively absent.32
Similarly, some evidence shows that when firms use their contracts as part
of their branding, they shy away from all-caps, suggesting that firms
themselves do not consider this method effective.33
Future discussions in disclosure law should focus on better alternatives
to all-caps. Here, there is cause for optimism—we find that certain
interventions can have a large impact on consumer consent. However, we do
not advocate any specific policy, and our findings should only be interpreted
as undermining the theory of all-caps.
I. CONSPICUOUS DISCLOSURE AND ASYMMETRIC INFORMATION
A.The Problem of Uninformed Contracting
Contracts are based on consent.34 A recalcitrant problem in contract
law, however, is that few consumers actually read the fine print, thus
compromising their consent.35 Inattention to the fine print encourages firms
to offer inferior terms because these terms will cut costs while not impacting
31 See infra Part I.
32 See e.g., ALEXANDER HIAM, MARKETING FOR DUMMIES, at 133 (4th ed, 2014) (“[A]void
long stretches of copy set in all caps.”)
33 David A Hoffman, Relational Contracts of Adhesion, U. CHI. L. REV. 1395 (2018).
34 See Omri Ben-Shahar, CONTRACTS WITHOUT CONSENT: EXPLORING A NEW BASIS FOR
CONTRACTUAL LIABILITY, 152 U. PA. L. REV. 1829 (2004)
35 Ayres & Schwartz supra note 3.

[p. 12]
12 ALL-CAPS [DRAFT] [VOL. ---
demand.36 Worse, firms will have an incentive to actively make terms harder
to read—i.e., “shroud” them—even in competitive markets..37
To deal with consumer mistakes concerning the terms of their
transactions, scholars advance several strategies. The dominant approach is
the promotion of mandatory disclosures.38 As Professor Bar-Gill, one of the
drafters of the new Restatement of Consumer Contracts, argued: “disclosure
mandates should be one of the main regulatory responses to the problem of
consumer misperception.”39 Similarly, Professor Sunstein argues that
“[p]roperly designed disclosure requirements can significantly improve the
operation of markets, leading consumers to make more informed
decisions.”40 Proponents of disclosure often use the Truth in Lending Act
(TILA) as an exemplar of successful smart disclosure.41 On the other hand,
there is a growing movement that is disillusioned with the disclosure project.
Ben-Shahar and Schneider, two of the leaders of this camp, argue that
“Mandatory disclosure may be the most common and least successful
regulatory technique in American law.”42 They consider TILA to be a “sour
36 See, e.g., Ayres & Schwartz, supra note 3, at 563 (If consumers are uninformed, “the
seller has too little incentive to offer good contracts.”). Oren Bar-Gill, The Behavioral
Economics of Consumer Contracts, 92 MINN. L. REV. 749, 774 (2008); Ryan Bubb & Richard
H. Pildes, How Behavioral Economics Trims its Sails and Why, 127 HARV. L. REV. 1593,
1644 (2014). There is also some evidence that firms intentionally sabotage disclosure, to
exacerbate the problem. Willis, supra note 17, at 1322-1326.
37 Bar-Gill supra note 36, at 744; Xavier Gabaix and David Laibson, Shrouded Attributes,
Consumer Myopia, and Information Suppression in Competitive Markets, 121Q. J. Econ. 505,
510 (2006); OREN BAR-GILL, SEDUCTION BY CONTRACT: LAW, ECONOMICS, AND PSYCHOLOGY
IN CONSUMER MARKETS, 19 (2012).
38 See e.g., Alan Schwartz & Louis Wilde, Intervening in Markets on the Basis of Imperfect
Information: A Legal and Economic Analysis, 127 U. PA. L. REV. 630, at 673 (1978) (arguing
that the chief remedy for market failures due to asymmetric information should be: “to provide
consumers with comparative price and term information”)
39 Id.
40 Cass Sunstein, Empirically Informed Regulation, 78 U. CHI. L. REV. 1349, at 1356 (2011)
41 Epstein, supra note 6, at 125, 128.
42 Omri Ben-Shahar & Carl E. Schneider, MORE THAN YOU WANTED TO KNOW, 3 (2014)

[p. 13]
2019] ALL-CAPS [DRAFT] 13
accomplishment.”43 What should come instead is a matter of on-going
debate.44
While scholars are debating the desirability of disclosure, courts and
legislators have adopted what can be called a “conspicuousness policy.” The
idea is simple: make key parts of the contract salient. This way, one could
reduce the cognitive strain, reading time, and cost-ineffectiveness of reading
the fine print. To encourage firms to use conspicuous disclosure, courts
condition the enforcement of certain key terms on their proper formatting.
So, for example, a disclaimer of the implied warranty under the UCC “must
be conspicuous” to be enforced.45 Similar requirements apply to disclaimers
of warranties under the Magnuson Moss Act, trial periods in consumer
contracts, disclosures of loans, and a variety of other contracts.46 Courts
also sometimes employ open-ended contractual doctrines, such as
unconscionability, unilateral mistake, and misrepresentation, to promote the
inclusion of conspicuous terms in the fine print.47
43 Omri Ben-Shahar & Carl E. Schneider, The Failure of Mandated Disclosure, 1
JERUSALEM REV. LEG. STUD. 83, at 86 (2015).
44 See e.g., Oren Bar-Gill, Defending (Smart) Disclosure: A Comment on More Than You
Wanted to Know, 11 JERUSALEM REV. LEG. STUD. 75–82 (2014) (arguing for simplified
disclosures); Willis (performance-based standards); Omri Ben-Shahar & Carl E. Schneider,
The Failure of Mandated Disclosure, 1 JERUSALEM REV. LEG. STUD. 83, 85 (2015) (reviewing
alternatives).
45 U.C.C. § 2-316(2); Melvin A Eisenberg, Disclosure in Contract Law, 91 CAL. L. REV.
1645, 1679 (2003).
46 15 U.S.C. § 2303 and 16 CFR 700.; CAL. BUS. & PROF. CODE § 17602(a); 15 U.S.C.
§ 1632 (Truth in Lending Act’s requirement that disclosure must be made “clearly and
conspicuously”). See also ALA. CODE § 8-19D-2(a) (“it shall be unlawful . . . [to imply in mail
solicitation] that the person being solicited has won . . . a prize or purported prize unless the
qualifying language appears in print that is clear, easily read, and conspicuous.”); K.S.A. 50903 (liability for failure to hold a sufficient quantity of a produce that is advertised as being
on sale, “unless the available amount is disclosed fully and conspicuously”); V.T.C.A., BUS.
& C. § 8.204; Dias v. Nationwide Life Ins. Co., 700 F. Supp. 2d 1204, 1216 (E.D. Cal. 2010);
Spray, Gould & Bowers v. Associated Internat. Ins. Co., 71 Cal.App.4th 1260, 1272, 84
Cal.Rptr.2d 552 (1999); Hadland v. NN Investors Life Ins. Co., 24 Cal.App.4th 1578, 1586,
30 Cal.Rptr.2d 88 (1994).
47 E. ALLAN FARNSWORTH, CONTRACTS, 248-49 (2004). § 211 R2K (“Where the other party
has reason to believe that the party manifesting such assent would not do so if he knew that
the writing contained a particular term, the term is not part of the agreement”).

[p. 14]
14 ALL-CAPS [DRAFT] [VOL. ---
While there is no general theory of what amounts to a conspicuous
display of information, all-caps play a dominant, and often dispositive, role
among both legislatures and courts.48 Various state laws explicitly mandate
that certain disclosures appear in all-caps.49 Other types of legislation simply
declare all-caps as an acceptable method of making text conspicuous.50
Courts, similarly, enforce terms only because they appear in all-caps.51 In
Rottner v. AVG, a consumer argued that software defect led to the loss of
information on his hard drive.52 The defendant argued that implied
warranties were disclaimed. The judge summarily noted that “Here, the
[contract] presents the disclaimer in capital letters in section 5c. . . . .
Consequently, Rottner's claim for any breach of the implied warranty will be
dismissed.”53
48 See e.g., FLA. ADMIN. CODE r. 2-18.002 (1996), MINN. STAT. ANN. § 559.21(3),
N.Y. GEN. BUS. LAW § 653(1), OHIO REV. CODE ANN. §3121.29 (mandating a block of 3
paragraphs of all-caps in child support orders); 18 DEL. ADMIN. CODE 1405-10.0 (2018); ALA.
CODE § 8-26B-10(c).
49 See, e.g., ALA. CODE § 8-26B-10(c). As noted, there is no generally accepted theory, and
some codes use forms without all-caps. See e.g., ALA. CODE § 8-25-2.
50 U.C.C. § 1-201(b)(10). Note that capitalization is not explicitly mentioned by the UCC
for the body of the text.
51 Sableman, supra note 13, at 24 (“courts have generally approved all-uppercase
treatments”); Beazley, supra note 3, at 8 (noting that “…all caps continues to be interpreted
as meeting the standard for ‘conspicuous type.’”); Willis, supra note 17, at 1349. Some
examples include Davis v. LaFontaine Motors, Inc., 719 N.W.2d 890, 895–96 (Mich. App.
2006); Doe v. SexSearch.com, 502 F. Supp. 2d 719 (N.D. Ohio 2007), aff'd on other
grounds, 551 F.3d 412 (6th Cir. 2008); Fleming Farms v. Dixie Ag Supply 631 So. 2d 922
(Ala. 1994); Karr-Bick Kitchens & Bath, Inc. v. Genini Coatings, Inc., 932 S.W.2d 877, 879
(Mo. Ct. App. 1996) (“The language excluding the warranties was written in capitalized
letters and was more prominent than the other type on the label. . . . The language thus
conformed with the definition of “‘conspicuous’”). Perlman 2012 WL 12854876, at *2 (S.D.
Fla. Apr. 3, 2012); Walnut Equip. Leasing Co. v Moreno (1994, La App 2d Cir) 643 So 2d
327; Boston Helicopter Charter, Inc. v Agusta Aviation Corp. (1991, DC Mass) 767 F Supp
363.; Potomac Plaza Terraces v QSC Prods., (1994, DC Dist Col) 868 F Supp 346, 26
UCCRS2d 1069.
52 Rottner v. AVG Techs. USA, Inc., 943 F. Supp. 2d 222 (D. Mass. 2013).
53 id at 232.

[p. 15]
2019] ALL-CAPS [DRAFT] 15
There are certain exceptions, but these mostly go to prove the rule.54 In
Herrera v. First Northern Savings and Loan Association,55 the tenth circuit
needed to decide whether an interest rate disclosure was “more conspicuous”
than other disclosures, as required by the Truth in Lending Act.56 The court
did not find that the APR disclosure met the standard, despite being in allcaps, because more than thirty other disclosures in the contract were also in
all-caps.57 Even in this decision and others like it, the court agreed that in
principle, all-caps is a mode of making text conspicuous.
Given the centrality of all-caps in legal practice and its social
importance, one would expect a large body of supporting evidence.
Strikingly, we could not locate any empirical support of this policy and only
scant theoretical justification.58 Instead, the evidence is mostly negative. In
a series of studies that started in 1928, psychologists generally found
negative effects of all-caps on reading speeds, slowing reading by as much
as 13%.59 One reason is that people are less experienced reading all-caps;
54 Bowdoin v. Showell Growers, Inc., 817 F.2d 1543 (11th Cir. 1987) (“By definition, a
post-sale disclaimer is not conspicuous in the full sense of that term because the reasonable
person against whom it is intended to operate could not have noticed it before consummation
of the transaction.”); but see Rinaldi v. Iomega Corp., No. 98C-09-064-RRC, 1999 WL
1442014, at *1 (Del. Super. Ct. Sept. 3, 1999) (finding that language was conspicuous even
though the terms were sent along with the packaged item) – or where the all-caps was on the
back of the page, see, e.g., Hunt v. Perkins Mach. Co., 352 Mass. 535, 541, 226 N.E.2d 228,
232 (1967) (“[T]he provisions on the back of the order cannot be said to be conspicuous
although printed in an adequate size and style of type.”); Sierra Diesel, 890 F.2d at 114
(finding capitalization on the back of the page was inconspicuous). But see Roger's Fence,
Inc. v. Abele Tractor and Equipment Co., Inc., 26 A.D.3d 788, 809 N.Y.S.2d 712 (4th Dep't
2006) (A clause may still be conspicuous even if on the back of the page and after the
transaction if there is a conspicuous notation on the front of the page directing attention to the
disclaimer on the back).
55 805 F.2d 896 (1986).
56 Id. at 898.
57 Id. at 900.
58 For a recent review, see Maria Lonsdale, Typographic Features of Text: Outcomes From
Research and Practice, 48 VISIBLE LANG. 29, 37-40 (2014). See also Willis, supra note 17,
at 1349 (noting the lack of supportive evidence).
59 See Miles A. Tinker & Donald G. Paterson, Influence of Type Form on Speed of Reading,
12 J. APPLIED PSYCHOL. 359 (1928). But see Jeremy J. Foster & Margaret Bruce, Reading
Upper and Lower Case on Viewdata, 13 APPL. ERGON. 145 (1982) (reviewing the evidence
and finding no negative effect of all-caps on reading speeds).

[p. 16]
16 ALL-CAPS [DRAFT] [VOL. ---
another is that all-caps letters lack ascenders and descenders,60 so that the
letters appear more homogenous.61 While instructive, this body of research
leaves much to be desired, as it is focused on non-legal texts and its main
finding—slower reading speeds—has ambiguous implications for consumer
law. In theory, slower reading could actually improve consumer consent, by
giving the consumer more time to reflect on the relevant term.
Among lawyers, all-caps is not commonly discussed—perhaps seeing
it as a mere technicality—but those who do, rarely endorse it. A leading
textbook on typography for lawyers counsels against the excessive use of allcaps.62 In a rare decision that adversely remarked on all-caps, Judge Kozinski
voiced a strong opposition: “there is nothing magical about capitals,” he said;
“Lawyers who think their caps lock keys are instant make conspicuous
buttons are deluded.”63
*
Courts and legislatures widely believe that all-caps makes a term
conspicuous, thus improving consumer consent. The literature review
reveals, however, that this belief has no empirical support. Although all-caps
exacts a heavy price from uninformed consumers by enforcing against them
especially onerous terms, it rests on speculation alone. We now set out to
present the first empirical evidence on all-caps in consumer contracts and
their effects on consumer consent.
60 Robbins, supra note 20, at 118-119.
61 See MATTHEW BUTTERICK, TYPOGRAPHY FOR LAWYERS, 202 (2012); Robbins, supra note
20.
62 See BUTTERICK, supra note 61. See also Robbins, supra note 20, at 116; Sableman, supra
note 13, at 9; Bryan A. Garner, Pay Attention to the Aesthetics of Your Pages, MICH. B. J.
(Mar. 2010), https://www.michbar.org/file/barjournal/article/documents/pdf4article1664.pdf.
Cheryl B. Preston, "Please Note: You Have Waived Everything": Can Notice Redeem Online
Contracts?, 64 AM. U. L. REV. 535, 553 (2015) (“Key sections in wrap contracts are
frequently presented in all capital letters, but that does not help.”); Beazley, supra note 3, at
2.
63 In re Bassett, 285 F.3d 882, 886 (9th Cir. 2002); OFFICE OF INV. EDUC. & ASSISTANCE,
U.S. SEC. & EXCH. COMM'N, A PLAIN ENGLISH HANDBOOK: HOW TO CREATE CLEAR SEC
DISCLOSURE DOCUMENTS 72 (1998) (proposing that text will not be written in all-caps).

[p. 17]
2019] ALL-CAPS [DRAFT] 17
II. ALL-CAPS IN ACTION: A STUDY OF INDUSTRY PRACTICES
Both casual observation and the caselaw suggest that all-caps
provisions are very common in consumer transactions.64 But how common
is very common? While we know that many consumer contracts are liberal
with their use of polysyllabic words and difficult, tortured grammatical
constructions, we know very little about their formatting.65 As
conspicuousness is ultimately a question of formatting, this gap in our
knowledge is troubling. Evaluating the practical importance of all-caps also
bears on our standard of proof for their effectiveness; all else equal, the more
prevalent they are, the more important it is to verify that they indeed achieve
their intended goals.
A.Methodology
To estimate the prevalence of all-caps in practice, we sought to examine
various types of common consumer contracts. We report here novel evidence
based on the analysis of the standard forms used by 500 of the most popular
websites.66 These forms serve the basis of hundreds of millions of individual
consumer contracts, as most US consumers have contractual relationships
with at least a few of these large firms.
64 See, e.g., Kelker v. Geneva-Roth Ventures, Inc., 303 P.3d 777, 783 (Mont. 2013) (finding
the arbitration clause in a payday loan agreement unconscionable because, inter alia, “no bold
or capital letters highlight[ed] the arbitration clause”); Mitsch v. General Motors Corp., 833
N.E.2d. 936, 940 (Ill. 2005) (finding the warranty of merchantability disclaimer required
under Magnuson-Moss act for the sale of used car conspicuous, even though it did not mention
merchantability, because it was “in all capital letters,” among other things).
65 Uri Benoliel & Shmuel I. Becher, The Duty to Read the Unreadable, 60 B.C. L. REV.
(forthcoming 2019); Michael Rustad & Thomas Koenig, Wolves of the World Wide Web:
Reforming Social Networks’ Contracting Practices, 49 WAKE FOREST L. REV. 1431, 1437
(2014).
66 The data was collected and generously shared by Uri Benoliel and Shmuel Becher and
forms the basis of their article Uri Benoliel & Shmuel I. Becher, The Duty to Read the
Unreadable 60 BOS. COLL. L. REV. (Forthcoming, 2019). The collection procedure is detailed
there.

[p. 18]
18 ALL-CAPS [DRAFT] [VOL. ---
The selection of the firms was made on the basis of the Alexa Top Sites
web service, which collects data on the most visited websites67 and is widely
considered to be a reliable measure.68 The sites in the sample include
household names such as Google, Facebook, Uber, and Amazon. The
contracts themselves are wrap contracts, which structure the relationship
between the firm and the consumer in relation to the usage of the website.
To analyze these contracts, we developed a script that algorithmically
detected the case of words, sentences, paragraphs, and headers.69 The script
counted all instances of a letter being capitalized, and attempted to classify
capitalization at the word, sentence, paragraph, and header level. One
challenge in this respect is that there is no unique way to identify headers–or
even paragraphs. The script defines a header as a sentence lacking a period.
Capitalization of a paragraph was defined as a paragraph containing over
80% of its content in uppercase.
B. Findings
Table 1 summarizes the main findings from the case analysis of the
contracts:
67 See Alexa Top Sites, AMAZON, https://aws.amazon.com/alexa-top-sites/ (last visited Mar.
1, 2019). The ranking itself is based on a combination of unique visitors and the number of
pageviews, per visitor. See How are Alexa’s traffic rankings determined?, AMAZON,
https://support.alexa.com/hc/en-us/articles/200449744 (last visited Mar. 9, 2019).
68 Joel R. Reidenberg et al., Disagreeable Privacy Policies: Mismatches Between Meaning
and Users' Understanding, 30 BERKELEY TECH. L.J. 39, 54 (2015) ("Alexa.com [is] the most
prominent measurement company for web traffic data."); Arjun Thakur et al., Quantitative
Measurement and Comparison of Effects of Various Search Engine Optimization Parameters
on Alexa Traffic Rank, 26 INT’L J. COMPUTER APPLICATIONS 15, 15 (2011);
("Alexa Traffic Rank is the most popular website traffic measurement unit").
69 The script uses Python’s library “Docx” which allows interaction with Word documents
and classifies words, sentences, and paragraphs.

[p. 19]
2019] ALL-CAPS [DRAFT] 19
Table 1: Analysis of Capitalization in the Standard Form
Contracts
As Table 1 shows, the great majority (~77%) of these contracts have at
least one paragraph that is fully capitalized. The use of capitalized headers is
also quite frequent, with 17.4% of all the headers formatted in all-caps.70
Contract drafters will also capitalize certain key terms and names, so we find
that roughly 9% of the words in these contracts are capitalized.
Overall, these findings demonstrate that capitalization is very common
in practice. In interpreting these results, it is important to bear in mind that
most American adults are a party to many of these contracts, which include
the contracts of firms such as Facebook, Amazon, Dell, and Uber. During
the collection of the contracts, these websites had 10 million unique
visitors.71 Hence, these 500 form contracts represent hundreds of millions of
individual contracts affecting the lives of most American adults.
Additionally, the use of capitalization in EULAs is not likely to be unique to
online contracts; if anything, the online format permits more formatting
opportunities than print contracts.72 Finally, it is remarkable how pervasive
all-caps are in legal texts relative to any other type of text. In marketing
70 Note, however, that there is no unique way to define headers and paragraphs, so this
estimate may be both under-and over-inclusive. We ran a verification analysis by hand and
found the script to be generally accurate.
71 See Benoliel & Becher, supra note 66.
72 See Sableman, supra note 13, at 9-10.

[p. 20]
20 ALL-CAPS [DRAFT] [VOL. ---
materials—where firms have a monetary incentive to increase
comprehension of their messaging—all-caps are rarely used.73
III. ALL-CAPS AND CONSUMER CONSENT: EXPERIMENTAL ANALYSIS
Considering the legal and practical importance of all-caps, it becomes
critical to know whether this mode of intervention in consumer contracts
succeeds in its stated goal of improving consumer consent. Testing the
effectiveness of the all-caps theory requires both a clear grasp of how
conspicuousness might improve the quality of consent and a clear
methodology that controls for the many potential confounders.
Courts have not expounded on why they believe all-caps improves
consent; instead, they summarily link all-caps to the prevention of surprise.74
Trying to trace the link opens a few possibilities. First, it is possible that
conspicuous language helps the consumer to economize attention. The
conspicuous formatting would indicate to the consumer where she should
spend most of her “attention budget,” because the terms are most important.
This possibility depends on contrast, so that conspicuousness is the quality
of the term’s visible difference from other parts of the text. Another
possibility is that conspicuous formatting improves the readability of the
text; a larger font type reduces eye strain and highlights letter structure or
simply draws attention more effectively.75 A third possibility is that all-caps
acts as a “fire siren”—it doesn’t make it easier to read or understand, but its
very existence alerts the consumer to the possibility that the contract is
especially onerous. A final possibility—and a counterintuitive one—is that
conspicuous language is helpful because it slows down reading speeds.76
This is potentially so because capitalized letters are homogenous and lack
what typographers call “ascenders” and “descenders,” or the parts of letters
73 See HIAM, supra note 32 (recommending that all-caps should not be used in marketing
materials)
74 Gatton v. T-Mobile USA, Inc., 152 Cal. App. 4th 571, 581, 61 Cal. Rptr. 3d 344, 352
(2007)
75 If this is what courts believe, one would expect them to require the capitalization of the
entire contract.
76 See supra note 25.

[p. 21]
2019] ALL-CAPS [DRAFT] 21
that go above (such as in “b”) and below (such as in “p”) the line of type.77
This homogeneity would tend to make reading slower, but also might have
the salutary effect of increasing the time consumers reflect on these terms.
Whichever of these possibilities is correct, they all point in the way of
a similar testable hypothesis: other things being equal, the consumer would
have better recall of the conspicuous term than if the term was
inconspicuous. As we have already noted the absence of any supporting
evidence, we should also note that there is a reason to suspect the
effectiveness of all-caps. As noted, all-caps letters are homogenous and lack
what typographers call “ascenders” and “descenders,” or the parts of letters
that go above (such as in “b”) and below (such as in “p”) the line of type.78
In addition, the capitalization of entire blocks of text makes the key terms
less conspicuous, as the conspicuousness of text may consist of contrast.79
Admittedly, one might hold a non-functional view of conspicuous
language. It is possible that courts think that posting a conspicuous sign is
enough to shift the burden to the consumer, or that they view all-caps as a
formality that serves other, non-consumer-oriented ends.80 Such theories,
however, have little in the way of support. Why should the mere act of
capitalization suddenly overcome the difficulty posed by consumers not
reading the fine print? If all-caps have no empirically discernible impact on
consumer consent, what normative force do they carry? And because we
could find no one making these arguments, much less justifying them, we
can restrict attention to the possibilities explored above, which relate
conspicuousness to informed consent.
Our position is that unless one can show that all-caps has a meaningful
impact on the quality of consent, all-caps should not be held to satisfy the
conspicuousness requirement at all. This is because the error cost of this
intervention—the enforcement of onerous but unknown terms on
consumers—can be very high. To bar a wrongful death lawsuit simply
because a clause in a contract was capitalized, one must have significant
77 See BUTTERICK, supra note 61, Robbins, supra note 20.
78 See BUTTERICK, supra note 61, Robbins, supra note 20.
79 On this view, low-caps would be conspicuous in a sea of all-caps text.
80 See generally Fuller, supra note Lon L. Fuller, Consideration and Form, 41 COLUM. L.
REV. 799, 800-801 (1941).

[p. 22]
22 ALL-CAPS [DRAFT] [VOL. ---
confidence that this method is indeed effective at improving consumer
consent.
A.Methodology
We are interested in seeing whether all-caps has any measurable impact
on consumer consent. The most direct measure would be the quality of
consumer consent “in the field,” but regrettably such an investigation
presents many difficulties and is fraught with a host of potential confounders.
To see whether the consumer read the contract at all one would have to
monitor the consumer closely from the early stages to the consummation of
the transaction. To evaluate whether the consumer’s understanding is due to
the contract or some other factor, one would also need to monitor the
consumer’s interactions with other consumers, the salespeople, or online
materials. There are also considerable variations in the way salespeople
communicate and treat different consumers,81 which could further confound
the analysis. These challenges make field research exceedingly difficult and
uncertain.
Randomized control trials, and in particular, lab experiments present a
rigorous method of evaluating the relevant factors. In the lab, it is possible
to control for all variation between the contracts, negotiations, and products.
Thus, when the researcher finds a variation in outcomes, he can attribute it
more directly to the treatment rather than some external factor.
We recruited American respondents through the popular online
platform Amazon’s Mechanical Turk (MTurk), a common staple of similar
work.82 This platform “has been studied extensively at this point. Its
advantages are that populations recruited via [MTurk] are more
representative of the national population than convenience samples (e.g.,
undergraduates) and that a variety of experimental findings have been
replicated using MTurk.”83 While not perfect, MTurk is a standard way of
81 See e.g., Ian Ayres, Fair Driving: Gender and Race Discrimination in Retail Car
Negotiations, 104 HARV. L. REV. 817 (1991) (finding, in a field experiment, that salespeople
offered worse terms to minorities)
82 See e.g., Furth-Matzkin & Sommers, supra note 21.
83 Tess Wilkinson-Ryan, The Perverse Behavioral Economics of Disclosing Standard
Terms, 103 CORNELL L. REV. 117, 150 n. 162 (2017) (internal citations omitted).

[p. 23]
2019] ALL-CAPS [DRAFT] 23
ensuring greater subject variability than the leading alternative of recruiting
undergraduates.84
We recruited, overall, 570 respondents. This sample size is larger than
comparable studies and power analysis shows that it is sufficiently large to
capture meaningful differences.85 The demographics of the sample, relative
to the general US population (in parentheses), are: 44.6% female (50.8%),
median age 38 (38), 75% white (60.4%), a median household income of
$52,000 ($57,652), and college degree or higher education 62.8% (30.9%).86
Relative to the general population, we find a general match, with the sample
skewing slightly male, white, and less wealthy. A robustness check did not
show any statistical differences along these dimensions. Nor did we have any
theoretical reason to expect that the race of participants will affect results in
any particular direction. A larger relevant skew is with respect to education,
although even here two points are worth remembering. First, this skew is
actually much smaller than that of common alternative recruitment methods,
most clearly, in undergraduate students.87 Moreover, some of this skew
would likely bias results in favor of all-caps, as more educated readers might
be, on average, more informed of the legal requirement to highlight key
terms in contracts using conspicuous language. Again, we did not find any
meaningful differences based on these factors.
Before delving into the description of the experimental design itself, it
is worth highlighting the basic challenge posed by testing the quality of
consent and our approach to overcoming it. Testing consent is not an easy
84 See generally Hillel J Bavli & Reagan Mozer, The Effects of Comparable Case Guidance
on Awards for Pain and Suffering and Punitive Damages: Evidence from a Randomized
Controlled Trial, 37 YALE L. POL’Y REV. 405, 453 (Citing “numerous studies” that show tht
“MTurk worker population is relatively representative of the general population—and
certainly more representative than traditional pools for surveys and experimentation”).
85 See supra note 21. The power analysis is based on the non-inferiority testing, as described
in Shein-Chung Chow et al., SAMPLE SIZE CALCULATIONS IN CLINICAL RESEARCH, 76-82 (3d
ed., 2018). Assuming proportions of 50% correct in both groups, a non-inferiority margin (𝛿)
of-0.1 and a sampling ratio of 1, the sample size for 𝛼=0.05,1−𝛽=0.95 is 538.
86 United States Census, https://www.census.gov/quickfacts/fact/table/US/LFE046217
(last visited July 31, 2019)
87 Joseph Henrich et al., The Weirdest People in the World?, 33 BEHAV. & BRAIN SCI. 61,at
63 (2010) (Finding that 67 percent of American subjects in psychology studies rely on college
students and that this population is often “at the extreme end of the distribution.”).

[p. 24]
24 ALL-CAPS [DRAFT] [VOL. ---
task, which may be the reason behind the paucity of research in this area.88
The key parameter of interest, in our estimation, is whether the consumer can
respond correctly to a question regarding the obligations they just incurred.89
However, in testing this, one runs into the problem that consumers may guess
based on background information they have from past exposure—rather than
engagement with the actual contract. Consequently, even if there is a
difference in the effect of different designs, it may be obscured by consumers
relying on past experience to respond rather than the contract itself. Our
novel solution to these problems, as developed below, was to draft an
arrangement that defies past expectations, presents multiple plausible
options, and is also sufficiently complex.
The design for this study uses a contract inspired by Spotify’s end
user license agreement.90 Such agreements are common among providers of
both online and offline services, who offer a free trial period that converts
automatically into a subscription-based service after the trial period lapses.91
Consumer agencies consider such agreements to have potential pernicious
effects due to their stickiness, as the consumer may unwittingly pay for an
unwanted subscription.92 Most courts and legislatures, however, are willing
to enforce such charges—so long as they are made in all-caps in the
consumer agreement—under the theory that such disclosure is
conspicuous.93 This study is a test, then, of whether the inclusion of such
clauses indeed improves the quality of consumer consent.
88 See supra notes 58-61 and accompanying text.
89 See discussion of this point see supra Introduction.
90 Spotify Terms and Conditions of Use, SPOTIFY, https://www.spotify.com/us/legal/enduser-agreement/ (last modified Feb. 7, 2019) (For an example of an automatic billing
disclosure, see § 3.3 of the Terms and Conditions of Use).
91 “Free” Trial Offers? FED. TRADE. COMM’N,
https://www.consumer.ftc.gov/articles/0101-free-trial-offers (last visited Feb. 9, 2019).
92 Koren Grinshpoon, License to Bill: The Validity of Coupling Automatic Subscription
Renewals with Free Trial Offers by Online Services, 28 FORDHAM INTELL. PROP. MEDIA &
ENT. L.J. 301, 303 (2018); “Free” Trial Offers?, supra note 91.
93 Grinshpoon, supra note 92, at 320–28 (Explaining that under California’s Automatic
Purchase Renewals Statute, for example, automatic billing terms must be disclosed “clearly
and conspicuously,” which is defined as, inter alia, “in larger type than the surrounding
text;”); 322 n.106 (listing many states that have adopted this requirement and definition). See
also Laura Koweler Marion and Leita Walker, Automatic Renewal Laws in all 50 States,

[p. 25]
2019] ALL-CAPS [DRAFT] 25
The respondents were told that they were simulating a free-trial sign up
for a new music streaming service called “TideTunes.” They were then given
and asked to read a two-page contract for the service, which consisted of 15
paragraphs. Respondents were asked to spend as much time reading this
contract as they would read any similar contract outside the experiment.
Subjects were randomly split among two groups, control and
treatment.94 In the former group, the entire contract appeared in low-caps,
i.e., normal formatting. In the treatment group, a test paragraph appeared in
all caps. The test paragraph for this study is as follows:
TERMS OF FREE TRIAL.
BY SIGNING UP FOR THIS FREE TRIAL, YOU ARE SIGNING UP FOR
MEMBERSHIP WITH TIDETUNES. YOUR MEMBERSHIP WILL CONTINUE
UNTIL YOU MANUALLY CANCEL IT. MEMBERSHIP INCLUDES AUTOMATIC
BILLING OF THE CARD WE HAVE ON FILE AT THE END OF THE MONTH FOR
THAT PERIOD. THE TERMS OF MEMBERSHIP APPLY TO THE FREE TRIAL. BY
PROVIDING YOUR PAYMENT DETAILS, YOU AGREE TO THE TERMS OF
AUTOMATIC BILLING. THE FREE TRIAL CANNOT BE TERMINATED PRIOR TO
THE END OF THE TRIAL. IF YOU DO NOT WISH TO BE CHARGED ON A
RECURRING MONTHLY BASIS, YOU MUST TERMINATE YOUR PAID
SUBSCRIPTION THROUGH YOUR USER ACCOUNT OR TERMINATE YOUR
ACCOUNT BEFORE THE END OF THE RECURRING MONTHLY PERIOD.
After being presented with the contract, respondents were moved to
a new page, from which they could not go back, and were asked: “Imagine
that you have signed up for a trial with TideTunes. When can you cancel
your trial?”. The options (presented in random order) were: (1) At any time;
(2) After the trial period (3) After seven days (4) After three months (5) After
fourteen days. The correct answer is number (2).
Before analyzing the responses, we should highlight that many studies
run the risk that respondents may use guesswork to respond to questions,
meaning that the responses are not affected by the stimuli presented to the
subject by the researcher. We used several measures to safeguard against this
risk.
Faegre, Baker, Daniels, available online at
https://www.faeghttps//www.faegrebd.com/webfiles/50State%20Survey%20Automatic%20Renewal%20Laws.pdfrebd.com/webfiles/50State%20Survey%20Automatic%20Renewal%20Laws.pdf
94 The covariates are well balanced between the two groups.

[p. 26]
26 ALL-CAPS [DRAFT] [VOL. ---
First, respondents on MTurk are incentivized to be attentive and
“[t]here is also evidence, both systematic and anecdotal, that Turk subjects
are particularly attentive, perhaps due to the formal mechanisms available
for giving them feedback that affect reputation ratings”.95 As a result, many
view this as a reliable tool of measurement.96
To enhance the quality of MTurk responses, we used a new special
service, called Positly, which adds a screening layer to MTurk.97 This service
allowed us to verify that all respondents were unique (i.e., that there was no
overlap between subjects in the studies), came from the US, and were within
the relevant age range. Importantly, the website uses several quality metrics
and attention and quality checks to screen out non-engaged users. Quite
tellingly, users on Positly are given an opportunity to respond to the survey,
and many complained that the content was boring and that it took them a
long time to slog through the entire contract.98
Third, we measured—behind the scenes—how long subjects spent on
reading the contracts. The average time to read (102 seconds) showed a nontrivial level of engagement with the text. Fourth, we presented subjects with
as many as five possibilities to choose from, in order to reduce the effect of
guesses. Finally, the fact that other experiments, reported below, produced
large differences also suggested that respondents were reacting to the stimuli.
95 Wilkinson-Ryan, supra note 83, at 150 n. 162.
96 On the reliability of MTurk, see Kristin Firth, David A. Hoffman, & Tess WilkinsonRyan, Law and Psychology Grows Up, Goes Online, and Replicates, J. EMPIRICAL LEG.
STUDIES (2017) (concluding that MTurk samples replicate well across testing platforms).
97 https://www.positly.com/participants/. Positly enhances the quality of respondents along
several dimensions: It aggregates data from independent researchers to screen out low-quality
participants; it conducts attentions checks; it screens duplicate responses by the same
individual; it uses a digital fingerprint technology to uniquely identify participants; and, it
uses IP addresses for geolocation. While none of these methods is perfect, it increases the
reliability of the baseline MTurk service and avoids some of its shortcomings.
98 Some complaints include: “[the contract] was a bit long and not that easy to answer the
main question without the agreement in front of me.”; “I was afraid I would have to return
this survey without pay since I couldn't remember certain verbiage from the contact.” “The
contract was difficult to understand”; “a lot of reading and it does not explain whether I was
right or wrong;” “[n]one of the contracts gave me enough time to read “ (with respect to Study
5). These complaints suggest that subjects were attentive and their complaints suggest that
they seriously attempted to respond to the questions at hand.

[p. 27]
2019] ALL-CAPS [DRAFT] 27
B. Findings
The main question of interest was how well consumers understand that
they can only cancel their trial after the end of the trial period. As noted, the
contract only permits the consumer to opt-out at the end of the trial period
(“THE FREE TRIAL CANNOT BE TERMINATED PRIOR TO THE END OF
THE TRIAL.”)
If all-caps improves noticeability and recall of hidden terms, we would
expect consumers in the all-caps group to answer this question correctly
more often than consumers in the low-caps group. Figure 1 summarizes the
findings.
Figure 1 Accuracy in All-Caps vs. Low-Caps
The key finding here is that respondents in the all-caps treatment failed
to show any improvement relative to the control. In fact, there were no
differences at all between the groups, and respondents in the all-caps group
were precisely as likely to respond correctly (or incorrectly) as respondents
in the low-caps group.
These findings allow us to reject, with high statistical significance, the
possibility that all-caps improves outcomes in a contractually meaningful

[p. 28]
28 ALL-CAPS [DRAFT] [VOL. ---
way.99 A non-inferiority test is a common method used to evaluate whether
one form of treatment is not worse than another. This is often applied in drug
trials, where the question is whether a new drug is at least as good as the drug
that is currently in use.100 By defining clinical significance, the researcher
can thus statistically evaluate whether the effects of the new drug are not
inferior to the current drug that is in use. Based on this method, we can test
whether low-caps is non-inferior to all-caps. Admittedly, there is no neutral
way to define contractual significance, but given the high error costs that we
noted—i.e., enforcement of especially onerous terms on the basis of the false
belief in their effectiveness—we believe that all-caps should be able to show
a meaningful improvement over low-caps before they should be approved.
If we adopt a ten percentage-point improvement benchmark (which indeed
may be too low for some), the data allows us to reject the hypothesis that
low-caps is worse than all-caps.101
It is important to observe that while we do not test for noticeability
directly, these findings bear on this issue. The test contract includes 15
paragraphs and remembering all of its content is not easy. If all-caps makes
text conspicuous, it should draw attention to its existence. Psychological
studies show that people tend to overly focus on salient features.102 We
would expect, then, that salience would reflect itself in better recall. The
failure of all-caps to improve on low-caps undermines the existence of a
positive notice effect.
The “fire siren” theory suggests that, even without reading, the
existence of all-caps would suggest to the consumer that the contract is
especially onerous. The data, however, allows us to reject this hypothesis.
99 To be clear, we do not conclude lack of effect on the basis of rejection of the null
hypothesis, but rather we test here the non-inferiority of the low-caps treatment.
100 See Chow et al., supra note 85, at 8. Gisela Tunes de Dilva et al., Methods for
Equibalence and Noninferiority Testing, 15 BIOLOGY OF BLOOD AND MARROW
TRANSPLANTATION 120 (2009).
101 With 𝛿=0.1, we can reject the hypothesis that 𝐻 :𝑝 −𝑝 <−𝛿 (where 𝑝 is the %
0 𝐿𝐶 𝐴𝐶
correct for each subscript category): 𝑧=2.85,𝑝<0.01. For proportion, this 𝛿 is equivalent
to having 138 instead of 125 correct responses in the all caps group, out of 283 participants.
For a lower 𝛿=0.05 we can reject the inferiority hypothesis with 𝑝=0.1.
102 See e.g., Joseph W. Alba & Amitava Chattopadhyay, Salience Effects in Brand Recall,
23 J. MKT. RES. 363 (1986)

[p. 29]
2019] ALL-CAPS [DRAFT] 29
The subjects were asked to answer a question with four potential answers.
The answers can be roughly ranked as being most lenient to the most
stringent, from cancellation at any time to cancellation after three months. If
the fire siren hypothesis were true, it would make subjects to opt for the
stricter options. But in fact, there were very few people—in each group—
who opted for either of the stricter options, and the great majority of people
chose one of the two more lax options. And in-between these two options,
all-caps respondents were less likely to choose the strictest one. Overall,
then, we do not find a fire-siren effect.
We then examine how age and all-caps interact. It is possible that the
all-caps intervention would provide value to certain age groups or that it
might harm others, as differences in generational norms, attention span,
eyesight, and so on might lead to different effects among age groups. To test
the age hypothesis, we estimated a logistic regression model where the
dependent variable was accuracy and the independent variable was age. We
controlled for race, education, and income.103 The following Figure reports
the results of the regression:
103 The results are unchanged even without the controls.

[p. 30]
30 ALL-CAPS [DRAFT] [VOL. ---
Figure 2 Average Marginal Effects of All-Caps with 99%
Confidence Intervals
The horizontal line is the benchmark, i.e., low-caps. The points show
changes in accuracy as a result of the all-caps treatment across different age
groups, ranging from 20 to 70. The bars around the points are the 99%
confidence intervals. As the figure shows, all-caps has a strong negative
effect on older readers.104 The older the reader, the more harmful the effect
all-caps has on their ability to answer the test question correctly. This is
notwithstanding a general trend in the data where older readers tended to be
significantly more likely to answer the questions accurately. To provide a
sense of the strength of this effect, the next figure splits the respondents into
two age groups:
104 𝑝<0.01. Note that for younger readers the apparent positive effect of all-caps lacks
statistical significance.

[p. 31]
2019] ALL-CAPS [DRAFT] 31
Figure 3 Percentage of Accuracy in Different Age Groups
80%
69%
70%
60%
50% 44%
42%
40%
40%
30%
20%
10%
0%
Below 55 Over 55
All Caps Low Caps
As Figure 3 illustrates, the difference in accuracy among younger
respondents is negligible. But for older audiences, the difference can be quite
stark. In respondents over 55, 60% were wrong in the all-caps group, relative
to only 31% in the low-caps group. This is a very large effect, effectively
doubling the error rate, and it is practically important given the stakes of
mistakes regarding all-caps clauses.
What might explain the tendency of older respondents to commit more
mistakes in the all-caps group than in the low-caps group? Impatience, lack
of motivation, and differential stakes of charges are all possibilities. An
additional explanation is that the use of all-caps is the formatting equivalent
of yelling or otherwise communicating anger.105 Thus, reading all-caps
would be an emotionally negative experience, which may lead older
respondents to avoid it more than younger respondents. What we find most
plausible is the explanation that all-caps impede reading because they
105 See Alice Robb, How Capital Letters Became Internet Code for Yelling, THE NEW
REPUBLIC (Apr. 17, 2014) https://newrepublic.com/article/117390/netiquette-capitalizationhow-caps-became-code-yelling; All Caps, PRACTICAL TYPOGRAPHY
https://practicaltypography.com/all-caps.html (last visited Feb. 9, 2019).

[p. 32]
32 ALL-CAPS [DRAFT] [VOL. ---
homogenize letter size, making it harder to distinguish between letters on the
basis of their ascenders and descenders.106
*
This study shows that the common practice of formatting certain
contractual terms in all-caps fails to improve outcomes for participants in a
meaningful way and that the practice in fact harms older readers. As we
emphasize throughout, the stakes of errors with the enforcement of all-caps
are high; these are some of the most consequential terms in consumer
contracts. Enforcing these clauses without evidence of their effectiveness
was always questionable; now we show positive evidence that this practice
is actually harmful. While caution is always prudent with lab experiments,
we believe that these findings are sufficiently clear to—at the very least—
shift the burden of proof. We will return to discuss these findings after
exploring some other aspects of all-caps.
IV. EXPLORING ALTERNATIVE JUSTIFICATIONS AND INTERVENTIONS
Our analysis so far has established that all-caps are very common in
practice, but that they lack any empirical support. Further, the evidence
presented here suggests that they fail to bear the burden of showing any
significant improvement over standard formatting. We now turn to a series
of exploratory studies that extend these results and test them under various
settings. We first check to see how all-caps performs under time pressure,
then we evaluate whether consumers may nonetheless show a preference for
all-caps, and finally, we look at whether some other modes of highlighting
text can be more helpful.
A.All-Caps under Time Pressure
1. Methodology
We have just seen that all-caps does not improve the quality of consent
in any meaningful way and impedes it among older readers. One limitation
of the primary experiment is the lack of any time limit. Subjects could spend
as much time as they saw fit on reading and reading the contract very closely
might diminish the usefulness of all-caps. In practice, however, time
106 BUTTERICK, supra note 61; Robbins, supra note 20.

[p. 33]
2019] ALL-CAPS [DRAFT] 33
pressures are ubiquitous, and one study found that as many as 65% of
respondents reported not reading the fine print because they were “in a
hurry.”107
Recall that under one theory, all-caps is useful in that it helps consumers
direct attention to the most important aspects of the transaction. Under this
theory, the positive effects of capitalization would be most noticeable under
time pressure, for then the consumer has to make an active choice where to
focus her attention. On the other hand, one might worry that if capitalization
results in text that is harder to read—a point we explore in the next study—
consumers may spend less time on this activity.
To test the effect of all-caps under time pressure, we designed an
exploratory series of three shorter contracts that were presented to readers
under a strict time limit. When reading the contract, the subject saw a timer
moving, noting the number of remaining seconds; once the time lapsed, the
subject was moved to the next page with the test questions. Each short
contract—described in the appendix—was followed by a multiple-choice
question that measured the reader’s recall of a specific term in the contract.
The term appeared in the test paragraph, which was either ordinary low-caps
(control) or all-caps (treatment). That is, the control group had no way of
knowing which paragraph contains the term paragraph, but the treatment
group could infer this on this basis of the use of all-caps used in this specific
paragraph alone.
We administered the test to 81 respondents, receiving 240 responses
overall (as there were three tests per respondents). The demographics of the
sample, (relative to the demographics of the general US population, in
brackets), are: 46% female (50.8%), median age 34 (38), 65% white (60.4%),
college degree or higher education 47% (30.9%), median household income
$50,000 ($57,652).108 The sample skews somewhat male, younger, white,
and poor, and significantly more educated. We do not have any theoretical
107 Robert Hillman, Online Consumer Standard Form Contracting Practices: A Survey and
Discussion of Legal Implications, in IS CONSUMER PROTECTION AN ANACHRONISM IN THE
INFORMATION ECONOMY?, 293 (2006) .
108 United States Census, https://www.census.gov/quickfacts/fact/table/US/LFE046217
(last visited July 31, 2019)

[p. 34]
34 ALL-CAPS [DRAFT] [VOL. ---
reason to expect this skew to point in any specific direction, but, coupled
with the small sample size, these findings should be interpreted with caution.
To determine what time limit to use, we first administered the test to a
small pilot group without a time limit. We measured the average time to read
for the test group and imposed an increasingly lower limit for each test.
Subjects were given 23 seconds to read test 1, 20 seconds to read test 2, and
only 15 seconds to read test 3. As can be seen by reviewing the contracts in
the appendix, these time limits are fairly challenging. The responses of the
pilot group were not included in the analysis.
2. Findings
Figure 3 summarizes our findings regarding the inaccuracy of responses
with the inclusion of the timer:
Figure 4 % of Mistakes under Time Pressure
75% 76% 74% 73%
63%
55%
As can be seen, subjects in the all-caps group failed to show any
improvement under time pressure. In fact, as we increased the time pressure
in tests 2 and 3, we see the low-caps group performing better, with slightly
higher accuracy rates, although only the second result approached statistical

[p. 35]
2019] ALL-CAPS [DRAFT] 35
significance.109 In Tests 1 and 3, respondents were also presented with the
option to respond “I don’t remember”. In both tests, the rates of failure to
remember were very similar—43.9% (low caps) vs 40% (all caps) in test 1
and 43% (low caps) and 43% (all caps) in test 3.
We see here, as in our primary experiment, that all-caps fails to improve
reader recall. The important feature of this variation was the use of a timer
with a strict deadline. The timer added both a physical and a psychological
constraint—reading long and complex texts within a short time is difficult
and the existence of a countdown timer can also impose stress. This is
arguably similar to a situation where the customer is reading a contract in the
dealership or at mortgage closing with the agent looking at them, expecting
them to sign the agreement. While lacking statistical significance, the results
are indicative that even under these fairly realistic constraints, all-caps does
not seem to improve outcomes.
The results of this study are noteworthy for those who believe all-caps
increases salience. If the use of all-caps is increasing the salience of the
text—indicating to the reader that this part of the text is not standard
boilerplate but rather an important part of the agreement—we would expect
readers to focus more attention on these clauses under time pressure. The
large text would indicate to them that this term, rather than any other, is
worth focusing one’s attention on. These initial findings, however, weigh
against the plausibility of the salience theory.
B. Subjective Sense of Difficulty & Reading Speeds
1. Methodology
What is the effect of all-caps on the consumer experience? Under one
theory noted above, capitalization helps consumers by increasing the font
size and, arguably, by using a typeface that is more cognitively efficient.
Unlike the theory of salience by contrast, this theory holds that capitalization
is important for making the text more accessible. If this theory is true, we
would expect at least one of the following hypotheses to be true. One,
consumers would tend to rate all-caps as easier to read and understand; two,
109 Based on noninferiority test, with 𝛿=0.1, the results of the hypothesis testing for the
three tests are, respectively, 𝑝 =0.25,𝑝 =0.1,𝑝 =0.2. Note that these results may be
1 2 3
related to the small sample size.

[p. 36]
36 ALL-CAPS [DRAFT] [VOL. ---
consumers would tend to spend less time reading a contract where the key
parts are effectively highlighted.
In the following exploratory variant of the study, we present 102
subjects with a version of the contract used for the primary study. The
demographics of the sample (relative to the demographics of the general US
population, in brackets), are: 45% female (50.8%), median age 36 (38), 84%
white (60.4%), median household income $56,277 ($57,652).110 This sample
skews considerably white, but otherwise has low skew. Again, this is an
exploratory study and it should be interpreted in this context.
Subjects were split among two groups, control and treatment. In the
control group, the entire contract appeared in low-caps. In the treatment
group, the contract was fully capitalized. Invisible to the participants, we set
a clock to measure the time from the moment the participant first saw the
contract until they clicked to the next page. Reading times were sufficiently
long to indicate engagement and we used attention checks and other quality
controls to verify engagement. 111 We also asked subjects to rate their own
sense of the difficulty of understanding the contract they read on a sliding
scale of 1-100, where 100 indicates the greatest difficulty.
2.Findings
The following Figure details the average ranking of the difficulty of
reading and understanding the text for subjects in both groups.
110 United States Census, https://www.census.gov/quickfacts/fact/table/US/LFE046217
(last visited July 31, 2019).
111 As a reminder, by low-caps we mean the standard English convention, with letters
opening a sentence and names being capitalized.

[p. 37]
2019] ALL-CAPS [DRAFT] 37
Figure 5 Self-Reported Difficulty of Reading &
Understanding the Contract
Panel A Panel B
Difficulty of Reading Difficulity of Understanding
70 70
59.8
54.6
60 60
49.2
48.2
50 50
40 40
30 30
20 20
10 10
0 0
All-Caps Low-Caps All-Caps Low-Caps
As the Figure shows, respondents rated reading and understanding the
capitalized contract as being considerably harder than respondents rated
reading and understanding of the low-caps contract. In terms of difficulty of
reading (Panel A), the capitalization treatment resulted in a rating of
difficulty that was roughly 22% harder. Understanding was also rated as
harder (Panel B), roughly 13% more in the capitalization group. The
difference in the difficulty of reading was statistically significant, suggesting
that all-caps did not make reading easier.112 The difference in the difficulty
112 𝑡(99)=2.088,𝑝<0.05

[p. 38]
38 ALL-CAPS [DRAFT] [VOL. ---
of understanding was not statistically significant (although it also was in the
same direction).113
This finding indicates that capitalization may result in a greater sense
of difficulty in reading the text and, to a lesser extent, understanding it. This
finding puts pressure on the theory that capitalization increases the
accessibility of legal texts, at least inasmuch as consumer preferences are
indicative of accessibility.
Some may doubt the validity of self-reported subjective rankings of
difficulty, but one should be cautious about dismissing this metric out of
hand; the negative valence of the experience of reading capitalized text—
whether or not it affects other metrics—may well dissuade consumers further
from reading contracts. It is also worth noting that consumers were not
ranking the contracts comparatively, i.e. not comparing the same contract to
another that is capitalized. Instead, the respondents reported their own sense
of difficulty regarding the single contract they saw. This suggests, in our
view, greater validity to the relative sense of confidence among the two
groups.
In terms of reading speeds, we found that members of the all-caps group
took longer to read the contract. The all-caps group averaged 94.7 seconds
relative to 83.4 seconds in the low-caps group. This difference (13%) was
not statistically significant, presumably due to the large variance in reading
times between members in each group or the smaller sample size.114 An
additional confounding factor is that members of the all-caps group, who
found the text more difficult to read, may have made less effort to read the
contract carefully.115 Still, it is remarkable that this is the exact same effect
size as previous work identified in non-legal contexts.116
We summarize this study as presenting early evidence against the
capitalization-as-accessibility theory. The capitalization of text resulted in a
113 𝑡(99)=1.21,𝑝=0.11
114 𝑡(99)= 0.78.𝑝=0.22
115 We note that in both groups, recall rates were similar, meaning that the increased reading
time did not result in higher likelihood to remember the content.
116 See Miles A. Tinker & Donald G. Paterson, Influence of Type Form on Speed of Reading,
12 J. APPLIED PSYCHOL. 359 (1928).

[p. 39]
2019] ALL-CAPS [DRAFT] 39
greater sense of difficulty reading the text and failed to improve the sense of
improved understanding of the text. Moreover, capitalization resulted in a
negative effect on reading times: not statistically significant, but potentially
large in practice. . The caveats presented above also apply here: Different
contracts may elicit different consumer responses, and it may be possible
that in other settings, consumers will not prefer a low-cap contracts, or that
some combination of formatting and content would make all-caps easier to
read. Still, our findings present the first empirical evidence on this issue and
they suggest the ineffectiveness of all-caps.
C.Taking the Con out of Conspicuous
1.Methodology
Our findings so far cast doubt on the idea that all-caps improves the
quality of consumer consent and suggest that, in some cases, all-caps
undermine it. In this study, we examine whether it is possible to improve the
quality of consumer consent through other means. Before we proceed to
describe this exploratory study, a preliminary comment is in order.
Designing communications is a difficult undertaking, conducted by
professionals who devote their careers to text design, marketing, and
copywriting. Our goal is not to argue that a single mode of communication
is always superior. Nor are we particularly interested here in discovering a
single mode of improving consent. Instead, we are interested in what
mathematicians sometimes call an “existence theorem;” i.e., discovering
whether it is possible, in principle, to improve contractual communications.
Such an inquiry is very timely, as many today are starting to abandon the
hope that consumers can read and understand contracts.117 If it is possible to
improve readability, perhaps not all hope is lost.
Overall, we recruited 241 respondents. The demographics of the sample
(relative to the demographics of the general US population, in brackets) are:
40% female (50.8%), median age 34 (38), 76% white (60.4%), median
117 See e.g., Omri Ben-Shahar & Carl E. Schneider, The Failure of Mandated Disclosure,
83 JERUSALEM REV. LEG. STUD. (2015); Ian Ayres & Alan Schwartz, The No-Reading
Problem in Consumer Contract Law, 66 STAN. L. REV 545 (2014).

[p. 40]
40 ALL-CAPS [DRAFT] [VOL. ---
household income $47,500 ($57,652).118 The sample skews somewhat male,
younger, and poorer, and significantly more white. We do not have any
theoretical reason to expect this skew to point in any specific direction, but
it is advisable to bear this in mind when interpreting our findings.
In this study, we presented respondents with a contract for the sale of
an RV, which included a liability disclaimer. The key paragraph, reproduced
below, was a disclaimer clause. The disclaimer waived liability for almost
all uses of the RV, but the seller assumed liability when the RV is driven on
the road. Respondents were allocated, randomly, to one of four groups,
illustrated in Figure 6 below. The control, as always, was the group where
the key paragraph was in low-caps. One treatment was all-caps. Another
treatment involved the use of a box, as suggested by some courts and
legislators, such as in the context of TILA.119 The last treatment was
“bold”—where we presented the contract in low-caps, but used boldface
formatting in a single key sentence. This treatment combines both boldface
and the selective highlighting of a single sentence.120 Note that given these
differences, this study is not a “horse-race” between boldface and
capitalization, because our concern is not with capitalization per-se, but with
blocks of capitalized text (i.e., all-caps).
118 United States Census, https://www.census.gov/quickfacts/fact/table/US/LFE046217
(last visited July 31, 2019).
119 Regulation Z, supra note 8. Bennett v. Matt Gay Chevrolet Oldsmobile, Inc., 408 S.E.2d
111, 114 (Ga. Ct. App. 1991).
120 As the last treatment involves changes—selective highlighting and boldface—it is not
possible to disentangle which of the two changes is more important. Our intention here,
however, is not to detect the best method of communication, but rather to see if any
interventions can be helpful. We leave the more nuanced analysis of design to future work.

[p. 41]
ALL-CAPS
Figure 6 % Four Design Choices

[p. 42]
ALL-CAPS
2. Findings
We measured the respondent’s answers to two test questions: whether
they can bring a lawsuit if the RV does not drive well off-road (the correct
answer is ‘No’), and whether they can bring a lawsuit if the RV does not
drive well on-road (the correct answer is ‘Yes’). The next figure describes
the error rates among the different interactions.
Figure 7 Error Rates, Four Treatment Groups
Low Caps
60%
57% All Caps
Box 52%
50% 48%
40%
All Caps
Bold 30%
30% 27% Box
Low Caps 24%
21%
20% Bold
13%
10%
0%
On-Road Off-Road
As these figures show, the bold treatment performed considerably better
than any other method of intervention. Focusing on the on-road question, the
use of bold text had a wrong answer rate of 27% relative to 57% (low-caps),
52% (all-caps), and 48% (box). In the off-road question, bold was again
associated with a low error rate (13%), followed again by low-caps (21%);
this time box did only marginally better (24%) and came ahead of all-caps
(30%).
To test the statistical significance of these differences, we estimated a
logistic regression model of the probability of accurately answering the

[p. 43]
2019] ALL-CAPS [DRAFT] 43
question with controls for the four different treatments. The following Figure
summarizes our findings:
Figure 8 Coefficient Estimate of Treatment Differences
Panel A: Off-Road Panel B: On-Road
In these figures, the horizontal line represents the baseline—low-caps—
and the bars the effectiveness of these interventions relative to this baseline
with a 95% confidence interval. As can be seen in Panel A, the Bold
treatment had a large, positive, and statistically significant in the off-road
question and a large, positive, but statistically insignificant effect in the onroad question (Panel B).121 The other treatments had a negative, but
statistically insignificant, effect relative to the baseline. The lack of a
statistically significant effect may well result from the absence of such a
difference, but also from the relatively small size of each of these groups.
These findings, first and foremost, support the possibility that some
methods of intervention can improve the ability of consumers to recall the
terms of their agreements. The success of the Bold approach suggests that
consumers readily react to communicative interventions, and their recall can
be significantly enhanced by designing interventions in a targeted manner.
This finding is consistent with early research done by psychologists who
121 Pr(𝑐𝑜𝑟𝑟𝑒𝑐𝑡)=𝐹(𝛽 +𝛽 𝐵𝑜𝑙𝑑+𝛽 𝐵𝑜𝑥+𝛽 𝐴𝐶+𝜖).𝑝<0.05.
0 1 2 3

[p. 44]
44 ALL-CAPS [DRAFT] [VOL. ---
found that readers prefer boldface over other types of emphasis.122 Still, our
goal here is not to design effective interventions. It is possible that other
variations would have been even more effective (including, perhaps, using
capitalization for just the key sentence). All we show here is that it is possible
that some well-designed interventions will have a large positive effect.
Another implication of this finding is that it validates the idea that subjects
in our studies are not engaging in guesswork, as their responses are sensitive
to the type of intervention.
Any optimism regarding the methods of intervention should be
tempered with the observation that other plausible interventions (all-caps and
box) failed to improve upon the benchmark of low-caps. These negative
findings highlight the difficulty of designing effective disclosure. Note,
however, that we cannot definitely say whether this is because these
interventions have no positive effect or because the difference did not
register given the sample size.
We noted above the fire-siren effect of all-caps and it is worth revisiting
it now. In the off-road question, the correct response was lack of a right; in
the on-road action, the correct response was that a right did exist. If the firesiren effect is real, we would expect the all-caps participants to believe that
a right does not exist in both cases at much higher rates than participants in
the low-caps group. The findings, again, cast doubt on the fire-siren effect,
as the response rates were fairly similar in both groups.
V. THE CASE AGAINST UPPERCASE
This paper studies one of the distinguishing markers of the legal genre:
The use of blocks of capitalized text known as all-caps. Courts and
legislators advance a deeply misguided policy whereby all-caps improve
consumer consent. Here we lay out the case against this policy and consider
several implications.
122 TINKER, supra note 18, 62.

[p. 45]
2019] ALL-CAPS [DRAFT] 45
A.Abolishing All-Caps
In the first part of the study, we showed the legal carte-blanche given to
all-caps. Courts are enforcing otherwise unenforceable terms because these
terms appear in all-caps. As such, consumers are locked into obligations in a
variety of contexts: wrongful death, liability for property damages,
arbitration agreements, waiver of implied warranties to name but a few.
Claire Donhau is a case in point.123 Her rock-climbing instructor gave her,
allegedly, wrongful instruction which led to her fall and the fracture of her
tibia in four different places. The Alaska Supreme Court upheld summary
judgment against her because the release form she signed “emphasized
language with simple words and capital letters.”124 As this case vividly
illustrates, courts will deny compensation from victims because of the allcaps formatting of the consumer contract—linking all-caps and consent.
Similarly, in a recent case, the Court of Appeals for the Eleventh Circuit
heard an appeal filed by homeowners against a shingles manufacturer that
sold allegedly defective and low-quality shingles that resulted in the early
deterioration of the homeowners' roofs. The homeowners wanted to file a
class-action, against the objection that an arbitration agreement that was
printed on the wrapper of the shingles prevented them from doing so. The
homeowners protested that they did not notice this provision, but the court
found persuasive the fact that the clause was written in all-caps.125 Thus, the
court denied the appeal and the homeowners were sent back to individually
arbitrate their cases.
Legislators not only permit the use of all-caps, they often mandate it. In
various settings, legislators require that certain disclosures will appear in allcaps; in others, legislators just list all-caps as a preferred mode of
disclosure.126 In either case, a firm that uses the statutory form immunizes
itself from later claims by consumers. Through this nexus of legislative and
123 Donahue v. Ledgends, Inc., 331 P.3d 342 (Alaska 2014)
124 Id.
125 Dye v. Tamko Bldg. Prod., Inc., 908 F.3d 675, 678 (11th Cir. 2018) (“[a]s particularly
relevant to this appeal, [the] limited warranty contains a mandatory-arbitration clause—
which, significantly, is also printed in its entirety, and in all caps, on the outside of every
shingle wrapper.”)
126 See supra Part I.B.

[p. 46]
46 ALL-CAPS [DRAFT] [VOL. ---
judicial policies, consumers are locked into some of the most onerous terms
for no other reason but their capitalization: Caps-Lock.
Firms react to this permissive legal environment in predictable ways.
In our analysis of the standard forms of 500 leading firms—forms which are
the basis of hundreds of millions of consumer contracts—we found that over
77% include at least one all-caps paragraph. This finding naturally leads one
to question the firms’ motives. Are firms naïve? Do firms genuinely believe
that using all-caps would promote consumer understanding? Or—worse—
do firms take advantage of the naïve judicial policy to hide some of the most
offensive, onerous, and costly terms in plain sight by using all-caps? The
latter option suggests a vicious dynamic. Not only do courts not protect
consumer interest by favoring all-caps, they invite abuse.
Our data cannot speak directly on this point, but we do think there is
some highly suggestive evidence that sheds light on these questions. The
legal context is but one of many where firms communicate with consumers.
When firms want to sell to consumers, they have every incentive to design
effective communications; indeed, this is the service provided by the multibillion dollar advertising industry. When looking at marketing materials, one
finds a rich, creative mix of text sizes, colors, typefaces, and backgrounds.
What one never finds is blocks of capitalized text, i.e., all-caps. Sure enough,
some individual words, and maybe even the occasional sentence, will be
capitalized. But blocks of homogenous capitalized text are all but absent.
This harkens back to the observation made at the outset; all-caps is a
hallmark of legal texts precisely because there is little reason to use all-caps
elsewhere. Moreover, there is some evidence that firms try to affirmatively
sabotage disclosure, making it less readable.127 Whatever is one’s view of
firms’ motivations, it should be clear that the legal system is permitting,
encouraging, and often outright mandating the use of all-caps.
It is against this background that our findings should be interpreted. The
primary experiment analyzed the responses of 570 people and demonstrated
that all-caps fails to improve consent within a reasonable margin of
effectiveness. Worse, the findings show that all-caps is harmful to older
readers. Readers over 55 were shown to understand their agreements
127 Willis, supra note 17, at 1322-26.

[p. 47]
2019] ALL-CAPS [DRAFT] 47
significantly worse when presented with all-caps text rather than standard
low-caps text. Importantly, our experimental design involved a simple
question—one that is relatively easy to answer correctly even on a quick
skim. Even with this simple metric, the older group answered incorrectly at
rates almost double than that of their same-age peers in the control group.
The findings also allow us to reject the “fire-siren” theory of all-caps.
Under this theory, all-caps is like a fire-siren in that one can easily hear it but
can hardly listen to it. As such, the very existence of all-caps would be a
signal of a contractually onerous term, even if the consumer does not
understand exactly what it might be. In fact, however, respondents did not
think however that the all-caps was more onerous than the low-caps one.
Even as a fire-siren, then, all-caps fails.
The interpretation of these findings can be informed by cognitive
research that suggests that the use of all-caps homogenizes the difference
between letter types, making it harder to read the fully capitalized text.128
The findings are also in line with common practical advice given by
lawyers.129 Another possibility is that the choice of typeface does more than
altering the form, but also changes the substance. Form, in language, is itself
a mode of communication. In the past, the usage of all-caps was meant to
designate “grandeur,” “pomposity,” or “aesthetic seriousness;” today, there
is a growing convention that all-caps is similar in effect to yelling.130 The
negative emotional valence associated with all-caps might make reading
more difficult or less appealing.
While exploratory in nature, this paper also tested the theory that allcaps would prove more beneficial in the presence of time-pressure. When
one has limited time, prioritizing attention becomes critical. If all-caps does
anything, performance under time-pressure would be the time for all-caps to
shine. We again could not detect any advantage provided by all-caps, but our
128 BUTTERICK, supra note 61; Robbins, supra note 20.
129 See BUTTERICK, supra note 61.
130 See Alice Robb, How Capital Letters Became Internet Code for Yelling, THE NEW
REPUBLIC (Apr. 17, 2014), https://newrepublic.com/article/117390/netiquette-capitalizationhow-caps-became-code-yelling; All Caps, PRACTICAL TYPOGRAPHY
https://practicaltypography.com/all-caps.html (last visited Feb. 9, 2019).

[p. 48]
48 ALL-CAPS [DRAFT] [VOL. ---
sample size in this specific experiment was fairly small and so our
conclusions are tentative.
It is also revealing that when people are asked to rate the difficulty of
reading, they rank all-caps as harder to read. Comparing the subjective
assessment of difficulty between individuals who read a contract containing
all-caps and those who read a contract in full low-caps, we found statistically
significant evidence that all-caps is harder to read. Individuals also rated the
all-caps contract as more difficult to understand, and although this finding
lacked statistical significance, it was in a similar direction—suggesting a
potential link between reading difficulty and understanding. Similarly, we
found evidence that reading times were longer under all-caps, but despite its
large magnitude (13%), this finding was not statistically significant. We
hypothesize that the longer reading times were counteracted by skimming,
as (presumably) subjects wanted to end the difficult experience faster.
Taken together, our empirical findings suggest the failure of caps-lock,
one of the most common and onerous consumer policies in the US.
*
We believe that there is a compelling reason to abolish judicial reliance
on caps-lock. Courts should no longer give any weight to the use of all-caps
in contracts. In fact, there may be a reason to treat all-caps with suspicion,
but we limit ourselves to calling for the renouncement of caps-lock.
In reaching this conclusion, we are well aware of the limitations of this,
or any other, lab study in terms of generalization, replicability, and external
validity. Our conclusion, however, rests on several mutually-enforcing
arguments that outweigh such concerns. First, our analysis of the literature
shows that the hypothesis that all-caps would improve consumer consent was
never validated; all-caps is instead an exercise in armchair theory. Courts
might have had a reason to think that all-caps could be effective, but resting
the full weight of such an onerous policy on an untested theory is deeply
misguided. Worse, the evidence in psychology that did exist at the time that
courts adopted this policy was negative: it showed that all-caps impeded
reading speeds.131 In evaluating all-caps, then, our starting position should
131 See Tinker & Paterson supra note 25.

[p. 49]
2019] ALL-CAPS [DRAFT] 49
be the general skepticism about any intervention that is supposed to easily
and dramatically increase the level of consumer consent.
Second, our findings suggest the practical failure of all-caps in legal
texts. Not only is all-caps not improving consumer consent, it actively harms
older audiences who in some settings may be the most vulnerable.
Consumers could identify their obligations no better under all-caps than
under normal print—and older readers did much worse. In light of this, it is
not surprising to find a consumer dislike of all-caps. Our evaluation of
subjective sense of difficulty, shows that individuals rank reading as much
harder when presented with text in all-caps. All-caps thus seems to be
violating the basic Hippocratic precept: first do no harm.
None of these weaknesses would have mattered much if the stakes were
low. But the stakes of error in this context are especially high. If a court
decides to enforce a liability waiver in the event of wrongful death because
the judge believes that putting the waiver in all-caps truly informed the
consumer, then all-caps has a series of unwanted effects. The consumer is
deprived of redress and compensation, which the consumer believed were
available to them. Indeed, the consumer may have even paid more under this
misguided belief and enforcing the waiver would deprive them of the benefit
of the bargain. From the firm’s perspective, the enforcement of the release
would leave a deterrence shortfall. For these reasons, the costs of error in
this context can be very high. Given that, to prove that the all-caps
intervention indeed improves consent, the bar should be set high. Exactly
how high is a matter of debate, but at the very least, we can agree that
speculation is an insufficient ground.
Fourth, we think there is a good a priori reason to approach all-caps
with great suspicion. As we noted above, the world around us is replete with
text that is meant to persuade consumers to buy products, text which is
designed by ingenious copywriters and shrewd advertisers. Yet, when firms
have a personal stake in the success of consumer communications, they
almost never employ large blocks of capitalized text in their brochures,
advertisements, and flyers. When these firms want to make attractive
features conspicuous, they use myriad design choices that have no
resemblance to the texts they use to obligate and bind consumers. As

[p. 50]
50 ALL-CAPS [DRAFT] [VOL. ---
Professor Hoffman showed, when firms have skin in the game, they can even
design fun and easy to read contracts.132
Taken together, then, we think the case against all-caps is compelling.
Of course, this study is not without limitations—the samples only
roughly represented the general US populations, we did not study many
possible formatting possibilities, did not test a large range of possible
contracts, and we were limited to responses in the lab. Still, we believe that
given the evidence presented here, courts and legislators should abandon the
preference given to all-caps. In the diverse contexts where conspicuousness
is required, courts should no longer accept all-caps as presumptively
conspicuous and thus retreat from a century-long jurisprudence in
disclaimers, waivers, arbitration clauses, choice of law provisions, and many
more. We are aware that legal traditions die hard. Yet, the stakes of this
specific legal tradition are extremely high and come at a severe cost to
consumers. If we care at all about informed consent, all-caps must be
abolished.
B.Stairway to Haven
Not all that is capitalized is conspicuous. Today, courts provide an
effective safe haven to firms that employ all-caps in their contracts. We
explained some of the dangers inherent in this practice, as consumers are
bound by terms they find hard to read and understand, and it encourages their
usage, irrespective of the effects of all-caps on consumers. Hence, the use of
a save-haven for all-caps appears ill-advised.
One might think that perhaps a different safe-haven is warranted, a
different mode of highlighting that firms can simply use to ensure
enforcement of the fine print. However, our central findings regarding the
failure of all-caps do not augur well for alternative safe-havens that are built
on mechanical, bright-line rules. It is not clear that it is possible to create
hard rules with broad applicability for human communications, which is a
subtle, complex, and context-dependent practice. This is especially
problematic, as sophisticated parties may learn to manipulate safe havens to
132 David A Hoffman, Relational Contracts of Adhesion, U. CHI. L. REV. 1395 (2018).

[p. 51]
2019] ALL-CAPS [DRAFT] 51
their advantage and we already noted that some firms are strategically
making disclosures less readable .133
There is, however, a more optimistic lesson here. We studied four
potential interventions in consumer contracts, including all-caps, low-caps,
a box around the text, and the use of boldface to highlight a specific key
sentence. The effect of correctly designed interventions is quite striking.
Subjects in the boldface group gave responses that were highly accurate,
responding correctly to one of the questions 73% of the time, relative to only
48% in the all-caps group. The success of boldface matches the findings of
early research done by psychologists that demonstrates that readers prefer
boldface to other types of emphasis.134 A 73% accuracy on a question
involving a long legal text—especially one that employs legal concepts—is
quite remarkable. This finding suggests that interventions can be quite
impactful if they are targeted and well-designed.
Tempering this optimism is the difficulty of employing this intervention
on a large scale. It is sometimes very difficult to condense the key terms to a
single sentence, and repeated use of this technique may have quickly
diminishing returns. Most critically, it is not clear that firms will even have
an incentive to properly design their communications, a point we return to
soon. These considerations, in combination with the limitations of results
from any lab study, suggest caution.
Still, Congress, regulators, and the courts will sometimes sacrifice
accuracy in favor of the certainty of bright-line rules, so there may be
practical pressure to offer such rules.135 With this in mind, one possibility for
future save havens is that courts will clearly distinguish between salience of
the paragraph for purposes of notice and its formatting for purposes of
reading. To increase notice, parties may be able to use a variety of signals of
importance—and may even include capitalization of the heading (as the
UCC itself suggests).136 With respect to salience markers, courts should be
permissive. At the same time, these markers should not extend to the text
133 Willis, supra note 17, at 1322-26.
134 TINKER, supra note 18, 62.
135 See Willis, supra note 17, at 1348-49 (noting the preference for hard, mechanical rules
in consumer law)
136 UCC § 1-201(10).

[p. 52]
52 ALL-CAPS [DRAFT] [VOL. ---
itself. Low-caps outperforms all-caps, at least among older readers, and if
the consumer’s attention is drawn effectively to the term there is no need to
alter the shape of the text. We recognize here, however, that setting a box
around the text written in low-caps did not prove itself effective in our study.
This again underscores the difficulty of setting hard rules, the importance of
experimentation, and the necessity to provide firms with an incentive to
improve communications. In the next subpart we describe an approach that
may, over time, coalesce into practices that might offer more robust safe
havens.
C.The Future of Disclosure
We are at a special moment in the life of consumer law. The new Draft
of the Restatement of Consumer Contracts has led to a heated debate among
scholars on whether courts should enforce terms in the fine print.137 One
group of scholars believes that market pressures, the existence of an
informed minority, and reputational pressures would lead firms to offer
efficient terms, and therefore courts should enforce the boilerplate.138
Another group believes that fine print terms should be presumptively
unenforceable absent a showing of informed consent.139 The Reporters of the
Restatement have taken the intermediate view that courts should be
permissive in questions relating to contract formation, but at the same time,
less permissive with enforcing these terms, seeing fine print terms as
potentially procedurally unconscionable.140
137 See Adi Robertson, A Contentious Legal Debate Over User Agreements Has Been
Delayed After Elizabeth Warren Called It ‘Dangerous’, The Verge (May 22, 2019).
138 Alan Schwartz and Louis L. Wilde, Imperfect Information in Markets for Contract
Terms: The Examples of Warranties and Security Interests, 69 VA. L. REV. 1387, 1462 (1983).
139 Dee Pridgen, ALI’s Proposed Restatement of Consumer Contracts – Perpetuating a
Legal Fiction, Consumer Law & Policy Blog (June 8th, 2016); Levitin et al., The Faulty
Foundation of the Draft Restatement of Consumer Contracts, 36 YALE J. REG. 447, 450 (2019)
(Noting the existence of disagreement on the “normative approach” of the new restatement).
See also Letter to ALI Council, Reject Council Draft No. 5 of the Restatement of Consumer
Contracts (Sept. 19, 2018) (calling for greater policing of contractual terms by the courts).
https://consumerfed.org/wp-content/uploads/2019/01/letter-opposing-council-draftconsumer-contracts.pdf
140 RESTATEMENT OF CONSUMER CONTRACTS (Preliminary Draft No. 3, October 26, 2017),
77.

[p. 53]
2019] ALL-CAPS [DRAFT] 53
The findings here are relevant to both sides of this debate. The failure
of all-caps, the most prominent form of smart disclosure, supports skepticism
about the meaning of consent to the fine print. The conspicuousness policy
is built on the idea that it is possible to avert or mitigate some of the noreading problem by highlighting key terms. The consumer would read more,
it is thought, if reading was made accessible. In practice, however, reading
of all-caps seemingly takes longer, the subjective feeling of understanding
falls, and recall does not improve over standard print and actually falls for
older readers. If a leading form of smart disclosure is ineffective, the
justification for the enforcement of fine print terms founders.
On the other hand, the success of some interventions is also quite
encouraging. In particular, we draw optimism from the finding that the
majority of consumers understood a legal disclaimer when it was presented
in an accessible form (the “bold” intervention).141 This suggests two distinct
avenues for future research and policymaking. First, it is possible to
effectively highlight information in a way that improves retention and recall.
Second, it is possible, at least in lab settings, to reach arguably satisfactory
levels of consumer understanding, even with jargon and text filled
paragraphs. To the extent such findings carry over to the world beyond the
lab, they should inspire some optimism about the possibility of designing
better forms of disclosure. Of course, whether consumers would read more
accessible disclosure is an open question; further, whether consumers should
read, especially when they lack the power to negotiate terms, is a normative
question.
An alternative approach is the use of performance-based contracts, at
least as an intermediary process. Performance-based contracts draw from
Lauren Willis’ powerful proposal that various consumer laws should be
based on proof of their effectiveness.142 That is, courts and regulators should
141 Consumer advocates may care more about the uninformed minority than any strict
majority of readers, while others would note that if a sufficient number of consumers
understand contracts, market pressure would lead to more favorable terms for all. See
generally Schwartz & Wilde, supra note 138.
142 Willis., supra note 17. For earlier discussions of this idea, see Howard Beales et al., The
Efficient Regulation of Consumer Information, 24 J.L. & ECON. 491, 530 (1981) (arguing that

[p. 54]
54 ALL-CAPS [DRAFT] [VOL. ---
move their “focus from firms’ actions to the effects of those actions on
consumers.”143 Under a somewhat modified version of her proposal, the
application of a performance-based approach would entail that courts should
by default hold key terms in the fine print unenforceable unless the firm can
affirmatively substantiate its claim that the term was made conspicuous.144
While one might worry that this will involve a large expense or expenditure
of time, we draw some optimism from the limited budget allotted to the
Article at hand.145 The research budget allocated to academics cannot hope
to compete with that devoted to marketing. Firms routinely engage in large
market research, known as A/B testing, where they test the slightest
variations in their marketing communications, sometimes using complex
statistical models. These are models, budgets, and techniques that can easily
be channeled to support consumer communications.146
Performance-based conspicuousness standards have a few important
advantages. First and foremost, they channel some of the genius that powers
advertising to the copywriting and design of the fine print. “Comprehension
standards allow firms to bring the full force of Madison Avenue to consumer
education in a way that is not possible for the government.”147 Designing
disclosure is hard, and currently, firms have very little motivation to do so.
Performance-based standards thus give firms some stake in informed
consent, because if their key terms cannot be shown to be effectively
communicated, those terms will not be enforced.
firms are best situated to design communications than regulators). See also Jeff Sovern,
Preventing Future Economic Crises through Consumer Protection Law or How the Truth in
Lending Act Failed the Subprime Borrowers, 71 OHIO ST. L.J. 761, 821 (2010) (suggesting
that lenders should affirmatively demonstrate that “a significant proportion of their borrowers
understood the terms of their loans"); M. Ryan Calo, Against Notice Skepticism in Privacy
(and Elsewhere), 87 NOTRE DAME L. REV. 1027, 1067 (2012) (proposing comprehension
standards for privacy disclosures.).
143 Willis, supra note 17, at 1314.
144 Here, conspicuous may be understood more broadly, as according with the consumer’s
expectation rather than comprehensible in isolation.
145 See also Willis, supra note 17, at 1366 (arguing that “inexpensive, painless, objective
testing of consumer factual knowledge could be surprisingly powerful”).
146 Ron Kohavi and Stefan Thomke, The Surprising Power of Online Experiments, HARV.
BUS. REV. (2017).
147 Willis, supra note 17, at 1337.

[p. 55]
2019] ALL-CAPS [DRAFT] 55
Second, and relatedly, all-caps safe havens have been stifling muchneeded creativity in this area. It is quite striking how little the informed
consent technology has progressed since the days of the typewriters. The
safe-haven approach is arguably the bottleneck and removing it could lead
to much-needed innovation. A third point builds on this insight. It is not
improbable that effectiveness-based standards are only an interim measure.
It is quite possible—almost inevitable—that best practices would quickly
evolve once firms have skin in the disclosure game. After all, the marketing
industry has certainly developed standards and common practices.
Given the goals and limits of this Article, the full case for performancebased conspicuousness will have to wait for another day. But we believe that
our findings here should support this project: It is quite possible to
dramatically increase consumer comprehension, although the design of such
interventions is not trivial and involves experimentation and cost. The future
of disclosure depends on engagement with the ideas of performance-based
conspicuousness.
Finally, we should emphasize that the effects of disclosure are
heterogeneous and one must be highly cognizant of their effect on vulnerable
groups. In our primary experiment, we highlighted how all-caps especially
harms older respondents. This is an important conclusion to bear in mind
when thinking about the future of disclosure, as the goal may not be to
maximize understanding across the board but may well be to minimize
misunderstanding across relevant parameters.

[p. 56]
56 ALL-CAPS [DRAFT] [VOL. ---
VI. CONCLUSION
An old anecdote tells of Niels Bohr, the Nobel-winning physicist,
whose door was adorned by a horseshoe. When asked by an incredulous
guest whether he believed in such superstition, Bohr replied that “I’ve been
told that it works even if you don’t believe in it.”148
This study explores the common practice of using all-caps in
consumer contracts and finds that the belief in their power borders on the
superstitious. Courts and legislators endorse this practice as a means of
improving consumer consent, given the lack of attention consumers pay to
the fine print. In reality, however, all-caps relies on no empirical support and
the evidence produced here suggests that all-caps is actively harmful to older
readers. The fact that all-caps is so widespread suggests that the stakes of
this superstition are significant even for those who do not believe in it. In
myriad cases, courts have been enforcing terms against consumers which
they erroneously thought consumers notice and understand.
Based on the evidence produced and collected here, we believe that
there is a robust case against uppercase. Courts should abandon their reliance
on all-caps as a proxy for quality consumer consent and consider other,
perhaps more contextual factors.
What may come next is best left to the genius of copywriters and the
prudence of lawyers. That courts have given a safe haven to firms that use
all-caps has stalled much innovation in this field, but there is great potential
for developments of new standards. As this Article demonstrated, the
targeted highlighting of key obligations has a strong and significant effect on
consumer consent. There are many other possible interventions, but none
will emerge so long as uppercase has the upper hand. We trust that this article
will help shift the burden of proof back to firms and help prevent future capslock.
148 I Understand It Brings You Luck, Whether You Believe in It or Not, QUOTE INVESTIGATOR
(Oct. 9, 2013), https://quoteinvestigator.com/2013/10/09/horseshoe-luck//.