Cornell

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

NOTE

THE EVIDENTIARY VALUE OF AUTOMATICALLY

TRANSCRIBED VOICEMAIL MESSAGES

George Cornell

I.! INTRODUCTION .............................................................................................. !

II.! AUTOMATICALLY TRANSCRIBED VOICEMAILS AS DISCOVERABLE,

ELECTRONICALLY STORED INFORMATION .................................................... !

A.! Federal Rule of Civil Procedure 34 ................................................... !

B.! The Lessons of Email and Voicemail E-Discovery ............................. !

1.! Email ............................................................................................ !

2.! Voicemail ..................................................................................... !

3.! Conclusion .................................................................................... !

III.! CHALLENGES TO THE ADMISSIBILITY OF AUTOMATICALLY

TRANSCRIBED VOICEMAILS .......................................................................... !

A.! The Rule against Hearsay .................................................................. !

1.! Examples of the Hearsay Rule’s Application to Email and

Voicemail ..................................................................................... !

2. ! Hearsay and Machine Statements: the Computer-Stored /

Computer-Generated Distinction ................................................. !

3.! Applying the Computer-Stored / Computer-Generated

Distinction to Automatically Transcribed Voicemails ................. !

4. !Getting Around the Hearsay Exclusion: Computer-

Generated Non-Hearsay ............................................................... !

5. !Getting Around the Hearsay Exclusion: Exceptions to the

Hearsay Rule ................................................................................ !

B.! The Requirement of Authentication or Identification ......................... !

1.! Federal Rule of Evidence 901 ...................................................... !

2.! Authentication by a Witness to the Conversation ........................ !

3. !Authentication where Witnesses to the Conversation Are

Unavailable: Identifying the Caller .............................................. !

4. !Authentication where Witnesses to the Conversation Are

Unavailable: Authenticating the Substance of the

Transcript ...................................................................................... !

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

C.! The Rule 403 Balancing Test: the Polygraph Analogy ...................... !

IV.! CONCLUSION ................................................................................................. !

I. INTRODUCTION

In the summer of 2009, Google Inc. began a limited offering of a new

service called Google Voice.

Google has since expanded the size of permitted

users, allowing existing users to invite friends and family.

Google Voice

utilizes customers’ existing home, cellular, and work phone numbers to

provide its users with a package of features that include cheap international

calling, free conference calling, the ability to consolidate multiple phone lines

into one Google Voice number, and free, automatic voicemail transcription.

When a Google Voice customer receives a voicemail, a completely automated

computer program converts the audio message into a text transcript, which the

service then sends to the user’s e-mail address or mobile device.

This

effectively allows a user to “search, sort, save, forward, copy and paste voice

mail messages.”

If the software is not very confident about how well it

transcribed certain words in a given message, the emailed transcript will

display those words in a lighter gray.

Google admits that while it expects the

quality of the transcriptions to improve as the software “gets smarter,” it is not

perfect.

Informal testing by one blogger revealed that the software struggles

under commonplace circumstances, such as when a nearby air conditioner

produces some background noise.

Under more challenging conditions, such

J.D., Boston University School of Law, Class of 2011.

Alex Pham, Apple Deaf to Google Voice App, L.A. TIMES, July 29, 2009, at B1.

Craig Walker & Vincent Paquet, Invite a Friend to Google Voice, GOOGLE VOICE BLOG

(Oct. 13 , 2009, 10:59 AM), http://googlevoiceblog.blogspot.com/2009/10/invite-friend-to-

google-voice.html.

David Pogue, One Number to Ring Them All, N.Y. TIMES, Mar. 12, 2009, at B1.

Id. at B8.

Id.

Basics: About Voicemail Transcriptions, GOOGLE VOICE,

http://www.google.com/support/voice/bin/answer.py?hl=en&answer=115986 (last visited

Apr. 4, 2010) (“The quality of the transcripts will vary depending on the caller, the

background noise, and whether the caller is using a microphone.”).

See David Gallagher, Help Us Test Google’s Hearing, GADGETWISE BLOG (June 26,

2009, 7:30 AM) http://gadgetwise.blogs.nytimes.com/2009/06/26/help-us-test-googles-

hearing/?hp.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

as in cases involving thick accents or nonsensical words from Lewis Carroll’s

Jabberwocky, the transcripts can be extremely inaccurate.

Despite these issues, automatic voicemail transcription stands to become

more prevalent in the coming years. In addition to Google, several other

companies already offer similar voicemail transcription services.

Some in

the communications industry predict that these services will be standard in the

near future.

In this age of “instant information gratification,” modern

professionals have little patience for the many steps it takes to hear and

respond to voicemails, especially when it is more efficient for them to read and

respond to text messages.

Users of this transcription technology appear to

include legal practitioners.

In fact, at least one transcription service has

actively marketed to law firms.

As the CEO of SimulScribe points out, “We

save law firms and other high-volume voice-mail users thousands of dollars by

eliminating wasted time listening to voice-mail.”

These uses suggest that

people will utilize voicemail transcription services in their professional and

personal capacities as the technology becomes more widely available.

Although Google only recently joined the ranks of companies already

See David Gallagher, Pushing the Limits of Google’s Speech Recognition, GADGETWISE

BLOG (June 29, 2009, 6:01 PM), http://gadgetwise.blogs.nytimes.com/2009/06/29/pushing-

the-limits-of-googles-speech-recognition/.

Jill Colvin, You’ve Got Voice Mail, But Do You Care?, N.Y. TIMES, Apr. 2, 2009, at

E1.

Id.

See Richard M. Georges, SimulScribe Signs Exclusive $17 Million Partnership

Agreement with Ditech Networks, FUTURELAWYER (Sept. 11, 2009),

http://futurelawyer.typepad.com/futurelawyer/2009/09/simulscribe-signs-exclusive-17-

million-partnership-agreement-with-ditech-networks.html (Blogger and practicing lawyer

Richard Georges discussing his use of Phonetag’s voicemail transcription service and noting

that “Phonetag’s transcription is . . . the main reason I haven’t had to suffer through

listening to a complete voice mail message in over a year. . . . And, if you do it like I do,

and have a transcription sent to you AND your secretary or assistant, you will have a written

record of the call for your file. Someday, someone is going to tell me that they never said

something, and I will pull out the written transcription.“).

See Dick Dahl, Free Court Decisions, Voice-Mail Text, Camera-Phone PDFs and

More, RHODE ISLAND LAWYERS WEEKLY, Apr. 7, 2008,

http://rilawyersweekly.com/blog/2008/04/07/free-court-decisions-voicemail-text-

cameraphone-pdfs-and-more/.

Id.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

offering voicemail transcription, it seems uniquely positioned to become a

dominant provider of such services.

In addition to being free and offering the

features already mentioned, Google Voice has “the potential to change the

rules of the game because of their ability to bring . . . people into their new

tools from their existing tools.”

These advantages have some observers

predicting hard times for competing transcription services.

Because Google

Voice seems likely to hold a significant market share relative to other

companies offering these services, the following discussion will treat Google

Voice’s completely automated voicemail transcription service as its model for

automatic voicemail transcription.

With every new communications medium, it seems inevitable that litigation

will eventually arise that involves, or even hinges on, information transmitted

through that medium.

Automatic voicemail transcription is somewhat of a

hybrid of two existing mediums: email and voicemail. Nevertheless, its

obvious function is still to memorialize information conveyed by the caller for

immediate or later review by the recipient of the call. In addition, at least one

user of this transcription technology retains such transcripts in contemplation

of future disputes over the messages’ contents.

Historically, companies that

have worked to prevent spoliation of evidence that may be important for future

litigation have focused more on retaining and preserving email than

voicemail.

This is due to factors including: (1) the storage capacity required

for retaining voicemail audio files, (2) the difficulty of searching voicemail

audio files for relevant information, and (3) the fact that often the voicemail

See Miguel Helft, Google’s Free Phone Manager Could Threaten a Variety of

Services, N.Y. TIMES, Mar. 12, 2009, at B9.

Id. (quoting Phil Wolff, editor of Skype Journal).

Pogue, supra note 3.

See, e.g., The Convoy’s Wheat, 70 U.S. 225, 230-31 (1865) (holding, in one of the

earlier federal cases involving a telegram, that the master of a vessel hired to deliver wheat

to a certain port should have telegraphed the consignees of the wheat in order to request

instructions after delivery to the intended port became impracticle, and faulting the master

for instead sending a demanding telegram only after deposting the cargo at an alternate

port); Ewan v. Tredegar Co., 88 F. 703, 704 (E.D. Va. 1882) (noting, shortly after the

invention of the telephone, that although a ship owner claimed to have notified the charterer

of the ship’s arrival by telephone, there was insufficient evidence to prove that the call

occurred, weakening the owner’s case against the charterer for demurrage because of

delays).

See Georges, supra note 13.

Steven C. Bennett, Voicemail, the next E-discovery challenge?, 14 PRAC. LITIGATOR

33, 38 (2003).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

audio was stored by the telephone carrier rather than on-site with the

company.

This tendency to preserve email over voicemail, along with the

increasing use of automatic voicemail transcription services, makes it plausible

that the only existing record of certain voicemails will be the automatically

generated transcripts residing on a company’s email server or a personal

computer. As early as 2003, practitioners foresaw that “[t]he ability to retain,

search, and easily transcribe voicemails, in theory, might make it possible to

use voicemail as a significant adjunct to (or in some regard, in lieu of) e-

mail.”

Now that technology has made automatic voicemail transcription a reality,

of what value are these transcripts to litigators? The following discussion will

examine major obstacles that litigators will face in attempting to use such

transcripts as evidence. With particular emphasis on federal law, I first

examine whether the Federal Rules of Civil Procedure definition of

electronically stored information is broad enough to allow parties to request or

compel production of automatically generated voicemail transcripts. I

conclude that the Rules’ language and case law trends support including these

transcripts among the types of discoverable, electronically stored information.

The remainder of the discussion will then focus on the problems that litigators

will face in getting such transcripts admitted as evidence. With primary focus

on the Federal Rules of Evidence, I will analyze three concepts that pose

peculiar challenges to any proponent of automatically generated voicemail

transcript evidence: (1) the rule against hearsay, (2) the requirement of

authentication or identification, and (3) Federal Rule of Evidence 403’s

exclusion of evidence that is low in probative value and high in prejudicial

effect. I conclude that these transcripts are susceptible to classification as

hearsay. Similarly, problems with the accuracy of these transcriptions may

also prevent authentication in many cases. Finally, Federal Rule of Evidence

403 may exclude some automatic voicemail transcription from evidence.

However, after examining analogous challenges to the admission of email,

telephone, polygraph, and computer generated evidence, I argue that none of

these challenges should operate as a per se bar to admissibility.

Id. at 35-36, 38.

Id. at 38.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

II. AUTOMATICALLY TRANSCRIBED VOICEMAILS AS DISCOVERABLE,

ELECTRONICALLY STORED INFORMATION

A. Federal Rule of Civil Procedure 34

For cases in which the Federal Rules of Civil Procedure control, whether or

not automatically transcribed voicemails should be subject to discovery

requires consideration of whether Rule 34 covers such transcripts. Rule 34

Provides in relevant part:

(a) In General. A party may serve on any other party a request within the

scope of Rule 26(b):

(1) to produce and permit the requesting party or its representative to

inspect, copy, test, or sample the following items in the responding

party’s possession, custody, or control:

(A) any designated documents or electronically stored information -

including writings, drawings, graphs, charts, photographs, sound

recordings, images, and other data or data compilations - stored in

any medium from which information can be obtained either directly

or, if necessary, after translation by the responding party into a

reasonably usable form; . . . .

As originally drafted, Rule 34 “focused on discovery of ‘documents’ and

‘things,’” and eventually was revised to include discovery of data

compilations.

Even before the 2006 Amendment to Rule 34 added the words

“electronically stored information,” the legal community interpreted the word

“documents” to encompass electronically stored information like email

“because it was obviously improper to allow a party to evade discovery

obligations on the basis that the label had not kept pace with changes in

information technology.”

In adopting the broad language of “electronically

stored information,” the advisory committee acknowledged that “[t]he wide

variety of computer systems currently in use, and the rapidity of technological

change, counsel against a limiting or precise definition of electronically stored

information.”

The advisory committee then stated, “Rule 34(a)(1) is

expansive and includes any type of information that is stored electronically. . . .

[It] is intended to be broad enough to cover all current types of computer-based

information, and flexible enough to encompass future changes and

FED. R. CIV. P. 34(a)(1)(A).

FED. R. CIV. P. 34 advisory committee’s notes to 2006 Amendment.

Id.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

developments.”

This expansive language and explicit intent to encompass

future innovations suggests that Rule 34 should be read to include

electronically stored transcripts of voicemails.

Just because such transcripts probably fall within Rule 34’s broad category

of electronically stored information does not necessarily mean that any given

transcript should be produced. As the advisory committee noted, “whether

material that falls within this term should be produced, and in what form, are

separate questions that must be addressed under Rules 26(b), 26(c), and

34(b).”

None of these rules, however, seems to present any problems that are

peculiar to voicemail transcripts as a class of electronically stored information.

Although the admissibility of such transcripts may be unclear, Rule 26(b)

specifically provides that “[r]elevant information need not be admissible at the

trial if the discovery appears reasonably calculated to lead to the discovery of

admissible evidence.”

Rule 26(b)(2)(B) states that “[a] party need not

provide discovery of electronically stored information from sources that the

party identifies as not reasonably accessible because of undue burden or

cost.”

Because Google Voice transcripts are delivered via email and are

likely stored along with other regularly delivered email, it is difficult to see

how retrieving those particular emails that contain transcripts would be

categorically more burdensome than retrieving other email. Finally, Rule

34(b) specifies the procedure for making and responding to discovery requests,

and does not include any language that would act as a blanket bar to transcripts

of voicemails.

Thus, the Federal Rules of Civil Procedure and the discovery

process these Rules established seem to encompass the transcripts produced by

automatic voicemail transcription services, and do not act as a per se bar to the

discovery of such documents.

B. The Lessons of Email and Voicemail E-Discovery

1. Email

Automatically transcribed voicemail messages are, in many ways, hybrids of

email and voicemail, with the transcribed contents of a voicemail conveyed via

email.

A look at how these more traditional communications are handled

Id. (emphasis added).

Id.

FED. R. CIV. P. 26(b)(1).

FED. R. CIV. P. 26(b)(2)(B).

See FED. R. CIV. P. 34(b).

See Pogue, supra note 3.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

suggests that voicemail transcripts will be viable targets for e-discovery

requests. Provided that the information within them is sufficiently relevant,

courts have held that emails are just as subject to disclosure as are paper

documents.

Noting that “[b]road discovery is a cornerstone of the litigation

process contemplated by the Federal Rules of Civil Procedure. . . ,”

courts

have held that “this is true not only of electronic documents that are currently

in use, but also of documents that may have been deleted and now reside only

on backup disks.”

Because of the confidential nature of many emails and the fact that internet

service providers often make assurances as to the privacy of email, one might

expect strong privacy claims against forcing litigants to disclose emails.

Courts analyze this expectation of privacy with reference to the Fourth

Amendment and federal and state statutes regarding internet

communications.

Generally, Fourth Amendment privacy protections do not

apply to discovery orders in suits involving private litigants.

Nevertheless,

some courts have considered Fourth Amendment protections when judging the

reasonableness of discovery orders.

Regardless of whether privacy

expectations have any bearing on email’s susceptibility to disclosure

requirements, “[g]enerally, courts have found no reasonable expectation of

privacy in e-mail messages.”

While “[i]ndividuals generally possess a

reasonable expectation of privacy in their home computers. . . . [t]hey may not,

See Rowe Entm’t, Inc. v. William Morris Agency Inc., 205 F.R.D. 421, 428 (S.D.N.Y.

2002) (refering to paper documents that defendant previously produced to plaintiffs, the

court noted, “[T]hose documents are plainly pertinent to the plaintiffs’ claims. To the extent

that the defendants’ e-mails contain similar information, they are equally discoverable.

Electronic documents are no less subject to disclosure than paper records.”).

Zubulake v. UBS Warburg LLC, 217 F.R.D. 309, 311 (S.D.N.Y. 2003) (citing Jones v.

Goord, No. 95 Civ. 8026, 2002 WL 1007614, at *1 (S.D.N.Y. May 16, 2002)).

Zubulake, 217 F.R.D. at 317.

See United States. v. Maxwell, 45 M.J. 406, 417-19 (C.A.A.F. 1996) (noting that for

Fourth Amendment purposes, a transmitter of an email message enjoys a reasonable

expectation of privacy against its interception under certain circumstances, but that once the

email reaches the recipient, “the transmitter no longer controls its destiny.“).

103 AM. JUR. Trials 123, §15 (2009).

See Doe v. Senechal, 725 N.E.2d 225, 231 (Mass. 2000) (Plaintiff lacked authority for

the proposition that a judge’s discovery order could constitute an unreasonable search and

seizure in a civil suit between private clients).

Id. at 231 n.13.

Mitchell Waldman, Annotation, Expectation of Privacy in Internet Communications,

92 A.L.R.5TH 15, § 2[a] (2001).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

however, enjoy such an expectation of privacy in transmissions over the

Internet or e-mail that have already arrived at the recipient.”

Courts have

also been unwilling to find a reasonable expectation of privacy in relation to

emails sent and received using an employer’s office computer, particularly

when the employer discloses in advance that it reserves control over such

systems.

Although courts have, in some circumstances, found that

reasonable expectations in the privacy of emails do exist,

the finding depends

largely on the specific circumstances of a given case, such as the recipient of

the email or the type of email at issue.

Thus, case law indicates that privacy

concerns have not posed a general hindrance to compelled email discovery.

2. Voicemail

The audio of voicemails themselves is an increasingly sought-after target in

e-discovery.

“The capacity of such evidence to have a major impact on the

outcome of a litigation (either affecting the substantive result, or putting

pressure on a party to settle) means that voicemail is a natural target for

discovery by parties in hard-fought litigation.”

Particularly now that sound

recordings are included amongst the types of electronically stored information

listed in Rule 34(a),

“[t]hese previously untouched pieces of evidence are

now fully discoverable . . . . [and] now unmistakably fall under the same

constraints to pinpoint process and disclose as do other permutations of ESI

[electronically stored information].”

Technological advances in the storage

of voicemails mean that many companies now use systems that integrate

United States v. Lifshitz, 369 F.3d 173, 190 (2d Cir. 2004).

See Doe v. XYC Corp., 887 A.2d 1156, 1165-66 (N.J. Super. Ct. App. Div. 2005)

(citing United States v. Simons, 206 F.3d 392, 398 (4th Cir. 2000) (employee did not have a

reasonable expectation in the privacy of emails drafted and received on employer’s

computer, where employer’s email policy stated that all emails composed, sent or received

on company computers were the property of the employer, not the employee)).

Waldman, supra note 41, at § 3[a].

United States. v. Maxwell, 45 M.J. 406, 418-19 (C.A.A.F. 1996) (“Expectations of

privacy in e-mail transmissions depend in large part on the type of e-mail involved and the

intended recipient.”).

Christopher Danzig, Hearing Aid: Audio Files Have Entered the E-Discovery Arena,

Adding New Risks for In-House Lawyers, INSIDE COUNSEL, May 2009, at 44.

Bennett, supra note 21.

FED. R. CIV. P. 34(a)(1)(A).

Michael Swarz, Voicemail, Web Conferences and Beyond: How Sound Recordings Are

Influencing the Way Corporate Counsel Conduct eDiscovery, 23 No. 13 CORP. COUNS.,

May, 2009, at 5.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

telephone and computer systems.

“These . . . systems sometimes use e-mail

notices that identify the caller, date, time, and duration of the call—which

provides the firm (or a litigation adversary) with the ability to trace an

employees’ receipt of messages, and can lead to additional burdensome

discovery demands.”

Digital voicemail recordings that are accompanied by

email notification are more like emails than traditional voicemail, and identical

production obligations will probably apply to them during discovery.

Though commentators predict the inevitable rise of voicemail as a

commonplace spoil of e-discovery,

case law is only starting to reflect the

shift towards audio discovery. “[D]igital voicemail [e-discovery] disputes

have yet to play a prominent role in reported decisions. . . .”

Nevertheless,

even before the 2006 Amendment to Federal Rule of Civil Procedure 34,

courts indicated that “discoverable electronically stored data includes voice

mail.”

While disputes over voicemail in e-discovery do not feature

prominently in case law, more recent cases involving recorded phone

conversations may also impact future decisions.

In In re Seroquel, the

plaintiffs moved for discovery sanctions, accusing defendants of, among other

things, “purposeful sluggishness” in the production of documents and data

from certain custodians.

In granting plaintiffs’ motion, the district court was

Mark Sidoti & Paul Asfendis, Haunted by Voices?, L. TECH. NEWS, May, 2009, at 26

(col. 1).

Id.

See supra notes 46-52 and accompanying text.

Sidoti & Asfendis, supra note 50; see also Sasha K. Danna, Note, The Impact of

Electronic Discovery on Privilege and the Applicability of the Electronic Communications

Privacy Act, 38 LOY. L.A. L. REV. 1683, 1702 (2005) (“[F]ederal case law addressing

discovery disputes over electronically stored voice mail is extremely scant.”).

See supra note 24-28 and accompanying text.

Danna, supra note 54, at 1702 (citing Thompson v. U.S. Dept. of Hous. & Urban Dev.,

219 F.R.D. 93, 96 (D. Md. 2003) (“[T]he scope of what is included in the phrase “electronic

records” can be enormous, encompassing voice mail, e-mail, [and other information] . . .

.”)); Kleiner v. Burns, No. 00-2160-JWL, 2000 WL 1909470, at *4 (D. Kan. Dec. 22, 2000)

(“[C]omputerized data and other electronically-recorded information includes, but is not

limited to: voice mail messages and files, back-up voice mail files, e-mail messages and

[other information]”) (internal quotations omitted).

Sidoti and Asfendis, supra note 50, at 26 (citing In re Seroquel Prods. Liab. Litig., 244

F.R.D. 650 (M.D. Fla. 2007); E*Trade Secs. LLC v. Deutsche Bank AG, 230 F.R.D. 582

(D. Minn. 2005)).

In re Seroquel Prods. Liab. Litig., 244 F.R.D. 650, 661 (M.D. Fla. 2007).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

persuaded, in part, by the plaintiffs’ contention “that the custodial production

has a great deal of missing data, e.g., although [defendant] has a system to

deliver voicemail, faxes, and video into Outlook [email] inboxes, none has

been produced.”

In Stamps v. Encore Receivable Management, Inc., the

plaintiff debtor planned to introduce a voicemail from a collection agency that

she recorded as substantive evidence in a civil suit alleging violations of the

Fair Debt Collecting Act.

The plaintiff subsequently objected to the

defendant collection agency’s request to produce the voicemail recording, and

requested that the court issue an order to delay the production of the recording

until after depositions.

Citing the plaintiff’s intent to use the voicemail

substantively, the district court rejected the plaintiff’s request and ordered

speedy production of the voicemail recording.

3. Conclusion

These cases demonstrate that courts have begun to view voicemails

themselves as discoverable, electronically stored information in civil suits.

This view is supported by the broad language of Federal Rule of Civil

Procedure 34, which specifically includes “sound recordings” as a type of

electronically stored information that is subject to disclosure.

Case law and

the same broad language of Rule 34 also make it clear that courts view emails

as falling well within the types of discoverable electronically stored

information.

Privacy concerns have generally not prevented courts from

treating email as subject to disclosure requirements, although some courts have

considered privacy expectations when gauging the reasonableness of discovery

orders.

To the extent that they are accurate, automatically generated

voicemail transcripts are essentially voicemails in email form. It stands to

reason, then, that the same considerations that apply to discovery of both email

and voicemail also apply to these transcripts. Since case law indicates that

emails and voicemails are both subject to disclosure, and keeping in mind the

Advisory Committee’s intent that Rule 34’s language be “flexible enough to

encompass future changes and developments,”

it seems reasonable to infer

Id.

Stamps v. Encore Receivable Mgmt., Inc, 232 F.R.D. 419, 420-22 (N.D. Ga. 2005).

Id. at 422-23.

Id. at 423-24.

See supra notes 24-28, 48-49 and accompanying text.

See supra notes 34-36 and accompanying text.

See supra notes 37-45 and accompanying text.

See supra note 28 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

that automatically transcribed voicemails will be subject to disclosure in civil

actions in which they are relevant.

III. CHALLENGES TO THE ADMISSIBILITY OF AUTOMATICALLY TRANSCRIBED

VOICEMAILS

The value of automatically generated voicemail transcripts to litigators will

depend largely on whether the information contained therein can be admitted

as evidence at trial. The remainder of this discussion will focus on some of the

more challenging issues litigators will face in their attempts to get such

information admitted. Because the Federal Rules of Evidence govern the

majority of federal judicial proceedings and have been adopted, either in whole

or in part, by forty-two states and Puerto Rico,

the discussion will focus

primarily on those rules.

A. The Rule against Hearsay

The Federal Rules of Evidence define hearsay as “a statement, other than

one made by the declarant while testifying at the trial or hearing, offered in

evidence to prove the truth of the matter asserted.”

For the purpose of this

definition, a statement is “an oral or written assertion. . . of a person, if it is

intended by the person as an assertion.”

Hearsay is not admissible as

evidence, “except as provided by [the Federal Rules of Evidence] or by other

rules prescribed by the Supreme Court . . . .”

1. Examples of the Hearsay Rule’s Application to Email and Voicemail

Email evidence is susceptible to exclusion on hearsay grounds. In a

proceeding for a protection from abuse order, a trial court refused to admit

email evidence submitted by the respondent.

The email was allegedly

written by petitioner’s mother and seemed to reference petitioner’s drinking

problem.

In affirming the trial court’s decision to exclude the email, the

GEORGE FISHER, EVIDENCE 2-3 (Foundation Press 2d ed. 2008) (2002).

FED. R. EVID. 801(c); see also FISHER, supra note 67, at 365 (suggesting that this

definition of hearsay can be streamlined to read, “Hearsay is [an out-of-court] statement . . .

offered in evidence by a litigant to prove the truth of the matter asserted by the declarant.“)

(alterations and emphasis in original).

FED. R. EVID. 801(a).

FED. R. EVID. 802.

Hood-O’Hara v. Wills, 873 A.2d 757, 759 (Pa. Super. Ct. 2005).

Id.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

Pennsylvania Superior Court referred to the state’s version of Rule 801.

Because respondent (appellant above) sought to offer the email as proof of

petitioner’s (appellee’s) alleged drinking problem, the court held that the email

was an out of court statement offered for the truth of the matter asserted.

Thus, the email was inadmissible hearsay.

In Means v. Cullen, the plaintiff

was a mentally ill inmate at a secure program facility who brought a civil suit

against a psychologist at the facility, alleging deliberate indifference to his

medical needs.

Plaintiff alleged that, after he expressed suicidal desires,

defendant replied that no one would care if he died.

Defendant subsequently

denied this statement in an email, and the email was included in an

investigative report regarding the plaintiff’s complaints.

An issue the court

addressed was whether the email could be considered for the purposes of

resolving defendant’s motion for summary judgment.

After finding that

several exceptions to the rule against hearsay did not apply,

the court

concluded that the email was inadmissible hearsay that could not be considered

in resolving the motion.

Cases about hearsay challenges to the admissibility of voicemail messages

are rare. This is probably largely due to the lack of cases involving voicemail

in general.

In United States. v. Somerset, the defendant was charged with

two counts of telephone harassment.

Following his conviction, the defendant

claimed that the trial court erred in admitting evidence of voicemails that the

defendant left for the victim despite the victim’s demands that the defendant

Id. at 760 (citing PA.R.E. 801) (“Under the Rules of Evidence, an out of court

statement that is offered for the truth of the matter asserted is excluded as hearsay.”).

Id.

Means v. Cullen, 297 F. Supp. 2d 1148, 1150-51 (W.D. Wis. 2003).

Id. at 1151.

Id.

Id. at 1151-52.

See id. (finding, inter alia, that the email’s inclusion in an investigatory report did not

except it from hearsay objections, noting that “[a]lthough . . . certain investigative reports

are excepted from the evidentiary rule barring hearsay, Fed.R.Evid. 803(8), statements made

by third parties recorded in the report are hearsay within hearsay and are inadmissible unless

they qualify for their own exception or exclusion to the hearsay rule, Fed.R.Evid. 805.”).

Means v. Cullen, 297 F. Supp. 2d 1148, 1152 (W.D. Wis. 2003).

See Sidoti & Asfendis, supra note 50.

United States v. Somerset, No. 3:03po002, 2007 WL 3005746, at *1 (S.D. Ohio Oct.

12, 2007).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

stop calling her.

The defendant claimed that the evidence of the voicemails

constituted hearsay evidence.

The district court rejected this argument,

noting that “the messages left by the Defendant qualify as admissions by a

party opponent and, therefore, do not constitute hearsay.”

The district court

also noted the trial court’s finding that the voicemail did not constitute hearsay

because the contents of the voicemail were “not offered ‘to prove the truth of

the matter asserted.’”

The district court does not elaborate on the trial court’s

reasoning.

However, since the telephone harassment statute cited in the

decision simply forbade the defendant from making a call to someone who had

previously requested not to be contacted,

the truth of the substance of the call

was not at issue. Therefore, the government probably offered the voicemail

evidence, not to prove the truth of any matter asserted therein, but rather to

show that the calls occurred at all. Though the Somerset court determined that

the content of the voicemails did not constitute hearsay under these facts, its

holding likely would have been different if the voicemail had been offered to

prove the truth of a matter asserted and if the caller could not be identified as

the party opponent.

Voicemail evidence also featured prominently in the White v. State murder

case.

Following the victim’s disappearance, her family members left her

several voicemails in which they pleaded for her to return their calls.

appeal from a conviction for murder, the defendant claimed that the contents of

the voicemails were inadmissible hearsay.

The court agreed with the

government that the messages were not hearsay “because they were not offered

Id. at *2, *5.

Id. at *5.

Id. at *5-6 (citing FED. R. EVID. 801(d)(2)).

Id. at *6 n.10.

Id.

United States v. Somerset, No. 3:03po002, 2007 WL 3005746, at *2 (S.D. Ohio Oct.

12, 2007).

See Denson v. State, 209 Ga. 355, 356 (Ga. 1952) (“A witness will not be permitted to

relate a conversation had with another person over the telephone, where such witness did

not know the other person or recognize his voice, and his identity is not established

otherwise than by what was said in the conversation itself. Such being hearsay evidence, is

inadmissible.”).

White v. State, No. 01-04-00410-CR, 2006 WL 727809, at *5-7 (Tex. App. Mar. 23,

2006).

Id. at *6-7.

Id.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

for the truth of the matter asserted therein. Instead . . . the messages were

offered merely to show that the calls had been made and to suggest that [the

victim], had she been able, would have responded . . . .”

Thus, the

voicemails did not fit the definition of hearsay specified by Texas’ version of

Rule 801.

The preceding cases demonstrate that courts analyze hearsay objections to

emails and voicemails in much the same way that they analyze the same

objections to paper and oral statements. In their analyses they treat both emails

and voicemail as statements which may or may not be offered for the truth of

the matters asserted therein. Since automatically generated voicemail

transcripts are a kind of hybrid of voicemail and email, and since their basic

function is to convert speech from one format into another, it seems likely that

courts will treat these transcripts as statements that can give rise to hearsay

objections.

2. Hearsay and Machine Statements: the Computer-Stored / Computer-

Generated Distinction

Putting aside the analogous treatments of email and voicemails, it is not

obvious that the rule against hearsay encompasses computer-generated

transcripts. At first glance, the Federal Rules of Evidence seem to exclude

such machine statements from possible sources of hearsay. Rule 801(a) limits

the definition of a statement to “(1) an oral or written assertion or (2)

nonverbal conduct of a person, if it is intended by the person as an

assertion.”

For the purposes of hearsay, the Rules define a declarant as “a

person who makes a statement.”

Therefore, “an objection on this [hearsay]

ground is ineffective against evidence of the output of machines. . ..”

However, though the output - in this case a transcript produced by a computer

program - may be the direct product of a machine, “information produced by

machines is, at one remove or many, a reflection of human design,

engineering, programming, calibration, and purposeful input, all aimed at

generating machine output.”

In cases where machine outputs are “a direct

and obvious reflection of human assertions or input, such as conversations

Id. at *7.

TEX. R. EVID. 801(d).

FED. R. EVID. 801(a) (emphasis added).

FED. R. EVID. 801(b).

4 CHRISTOPHER MUELLER & LAIRD KIRKPATRICK, FEDERAL EVIDENCE § 8:13 (3d ed.

2007).

Id.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

captured by recording devices, printouts of business data, and the content of

websites. . .ordinary hearsay analysis is appropriate.”

100

On the other hand,

[w]hen information from machines is mostly a product of mechanical

measurement or manipulation of data by well-accepted scientific or

mathematical techniques, the usual approach is to try to assure accuracy

by requiring the proponent to lay a proper foundation by showing that the

machine and its functions are reliable, that it was correctly adjusted or

calibrated, and that basic data put into the machine are accurate.

101

In an effort to differentiate between those computer outputs that should be

subject to hearsay objections and those that should not, some courts and

commentators employ a conceptualization of computer outputs that

distinguishes between “computer-generated data” and “computer-stored

data.”

102

Computer-stored data “represents only the by-product of a machine

operation which uses for its input ‘statements’ entered into the machine by out

of court declarants.”

103

Computer-generated data, in contrast, is the result “of

the computer’s internal operations. . . [and] does not represent the output of

statements placed into the computer by out of court declarants.”

104

Other

courts simply regard all computer records as hearsay, “admissible only under

the business records or public records exceptions.”

105

In Armstead, the

Louisiana Supreme Court refused to treat as hearsay certain automatically

generated telephone call logs that linked the defendant to obscene phone

calls.

106

The Armstead court acknowledged that “computer printouts which

reflect computer stored human statements are hearsay when introduced for the

truth of the matter asserted in the statements.”

107

However, the court

distinguished between such stored statements and the telephone logs at issue,

noting that “we are not dealing with computer stored human statements. . .the

evidence in this case was generated solely by the electrical and mechanical

100

Id.

101

Id.

102

See, e.g., State v. Armstead, 432 So. 2d 837, 839-40 (La. 1983); Adam Wolfson,

Note, “Electronic Fingerprints”: Doing Away with the Conception of Computer-Generated

Records as Hearsay, 104 MICH. L. REV. 151, 158-59 (2005).

103

Armstead, 432 So. 2d at 839.

104

Id. at 840 (listing as examples a seismograph’s recordings of geophysical occurrences

and a flight data recorder’s record of physical conditions onboard an aircraft).

105

Hawkins v. Cavalli, No. C 03-3668 PJH, 2006 WL 2724145, at *12 (N.D. Cal. Sept.

22, 2006) (citing Wolfson, supra note 102, at 155).

106

Armstead, 432 So. 2d at 840.

107

Id. at 839.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

operations of the computer and telephone equipment, and was not dependent

upon the observations and reporting of a human declarant.”

108

The Armstead

court noted that scholars had previously recognized the distinction between

computer-stored data and computer-generated data but, perhaps

acknowledging the new territory it was charting, lamented that “the writers

have not satisfactorily developed a consistent framework for evaluating

computer generated data.”

109

In one attempt to create such a framework, Adam Wolfson argues that

federal courts should adopt Louisiana’s Armstead approach of discerning

between computer-generated and computer-stored data.

110

Wolfson notes that

computer evidence often has a “highly unified presentation,” which causes

“judges and lawyers alike [to] miss the crucial distinctions that make parts of

the evidence admissible and other parts barred under the FRE and various

court precedents.”

111

For example, suppose that two individuals negotiate and

settle upon the terms of an agreement for certain services on an Internet forum

like a message board. Prior to participating in the forum, each human user

selected an alias or user name by which to identify themselves and their forum

postings. For every posting made in the course of the negotiations, each

person fills in a text field with his or her desired terms, representations, and

any other input that he or she decides to type. When the human user clicks to

upload his or her posting, the drafted text is submitted to the server handling

the online forum. Whenever somebody views the posting online, the server

automatically displays the time of the posting and the user name associated

with the posting along with the human-drafted text.

112

Suppose now that one or more of these postings becomes the subject of

litigation and that one of the parties seeks to introduce evidence that one of the

postings was made at a specific time by a certain user. The proponent wishes

to admit a printout from the forum showing the message text, timestamp, and

user name. Opposing counsel might then object that the document reflects an

assertion by the human user who created the post that (1) a person with a

certain user name (2) made certain claims (3) at a certain time. The opponent

would likely emphasize the human involvement in drafting the substantive text

108

Id. at 839-40.

109

Id. at 840 n.3.

110

Wolfson, supra note 102, at 158-61.

111

Id. at 168-69.

112

See, e.g., United States v. Hamilton, 413 F.3d 1138, 1142-43 (10th Cir. 2005)

(discussing how a computer automatically generated date and username information to

accompany the defendant user’s postings to a newsgroup).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

of the post and in triggering the post’s upload. The judge, confronted with a

single document conveying all this information, may view the entire

submission as inadmissible hearsay. However, the unified presentation of this

information, together on one printed web page, belies the fact that the

information is really an amalgamation of computer-generated and computer-

stored information.

113

While the text portion of the posting that was drafted by

a human is easily categorized as computer-stored information because it

“represents. . .the by-product of a machine operation which uses for its input

‘statements’ entered into the machine by [an] out of court [declarant],”

114

the

same is not true for the timestamp. In this example, the timestamp is an

automatic addition to the post by the server that involved no human input other

than the triggering of the upload. It is therefore a computer-generated piece of

information that fails to meet the definition of a statement for the purposes of

hearsay.

115

This hypothetical illustrates the need for judges to be able to

differentiate between those portions of seemingly unified computer outputs

that are computer-stored statements and those that are computer-generated.

116

It is unclear exactly where federal courts draw the line between outputs that

are primarily the statements of a machine and outputs that are essentially

human assertions, which are subject to hearsay objections.

117

However, in the

113

See Wolfson, supra note 102, at 167-69 (arguing that this unified presentation causes

judges and lawyers to conflate computer-stored information that may be hearsay with the

computer-generated “electronic fingerprints” that accompany such information.).

114

See State v. Armstead, 432 So. 2d 837, 839 (La. 1983) (defining “computer-stored”

information).

115

See supra notes 96-98 and accompanying text.

116

See Wolfson, supra note 102, at 167-69 (advocating a three part test for analyzing

computer outputs and noting the importance of breaking a computerized record “into its

constituent parts, which are usually a combination of computer-stored and computer-

generated data.”).

117

Compare Hamilton, 413 F.3d at 1142-43 (holding in a child pornography prosecution

that “header” information uploaded along with pornographic images, such as the user’s

screen name, the subject of the posting, and the posting date, was not hearsay because the

“header information was generated instantaneously by the computer without the assistance

or input of a person.”) with United States v. Cowley, 720 F.2d 1037, 1044 (9th Cir. 1983)

(holding that a machine-affixed postmark was hearsay because “a postal official is

responsible for setting the machine and causing the letters to pass through it . . . [t]he

postmark is thus the postal official’s written assertion that the letter passed through his

hands at the [particular] post office on a particular day.”). One could argue that the screen

name and subject information automatically affixed to each pornographic upload in

Hamilton was just as susceptible to initial human input as the settings on the postmark

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

computer context, case law suggests that the Armstead distinction between

computer-generated and computer-stored data is gaining recognition.

118

Evidentiary treatises also seem to recognize the distinction, though they do not

necessarily use the same terminology.

119

3. Applying the Computer-Stored / Computer-Generated Distinction to

Automatically Transcribed Voicemails

The distinction between computer-stored data and computer-generated data

has major implications for the evidentiary value of automatically transcribed

voicemails. The process by which the Google Voice program transcribes

voicemails into text is automatic, involving “no human effort.”

120

It is unclear

whether other transcription services can claim to have the same degree of

automation.

121

Nevertheless, even for those voicemail transcription services

that require no human aid in translating audio voicemails into text, transcribed

voicemails do seem to be obvious reflections of human assertions and input

because they are intended to represent oral statements of human callers.

122

The

form of the statements change, thanks to the program, but the fact that they are

assertions does not change. Since the output of automatic transcription

programs will usually use human assertions for input (i.e. the statements within

the voicemails), hearsay analysis will apply to many voicemail transcripts,

even though they are generated by a machine.

123

Thus, it seems that if a

proponent seeks to admit such a voicemail transcript to prove the truth of the

machine in Cowley.

118

See, e.g., Hamilton, 413 F.3d at 1142 n.4 (noting that its decision that data

accompanying uploaded pornographic images was admissable “might [have been] different

if ‘computer-stored’ data, as opposed to ‘computer-generated’ data, were involved.”) (citing

People v. Holowko, 486 N.E.2d 877, 878–79 (Ill. 1985)); Hawkins v. Cavalli, No. C 03-

3668 PJH, 2006 WL 2724145, at *11-12 (N.D. Cal. Sept. 22, 2006) (citing Armstead, 432

So. 2d at 840; Wolfson, supra note 102, at 157-58).

119

See 4 MUELLER & KIRKPATRICK, supra notes 100-01 and accompanying text.

120

See Pogue, supra note 3.

121

There has been some controversy over just how much human intervention has been

involved in some so-called “automatic” transcription services. See, e.g., Urmee Khan,

Security Fears over Voice-to-Text Service, THE DAILY TELEGRAPH, July 24, 2009, at 12

(discussing a BBC investigation into SpinVox, an automatic voicemail transcription service

provider that has admitted to doing at least some of its transcription through humans at call

centers in countries like Egypt and the Philippines).

122

See generally Pogue, supra note 3 (describing the function of Google Voice’s

automatic transcription service).

123

See 4 MUELLER & KIRKPATRICK, supra note 100 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

matter asserted therein, the submission will be susceptible to hearsay

objections.

124

4. Getting Around the Hearsay Exclusion: Computer-Generated Non-

Hearsay

Though automatically generated voicemail transcripts are susceptible to

hearsay objections, this does not serve as a general, per se bar to their

admission. As in the hypothetical about the message board postings,

125

Google Voice transcript is a combination of computer-stored information and

computer-generated information.

126

When a Google Voice subscriber receives

an automatically emailed copy of a voicemail transcript, the email contains: (1)

the phone number of the caller, (2) the name (if any) that the subscriber has

chosen to associate with that number, (3) the date and time of the call, and (4)

the text of the transcript itself.

127

While the text of the transcript itself reflects

the human input of the voicemail, the computer automatically generates the

caller’s number and the date and the time of the call.

128

The computer-

generated time and date information of the email are analogous to the

admissible timestamps that accompanied the newsgroup postings in

Hamilton.

129

Similarly, the automatically included telephone number is

analogous to evidence related to telephone traces and caller I.D. displays that

courts have viewed as computer-generated non-hearsay.

130

Therefore, the

124

See supra notes 68-108, 117-23 and accompanying text.

125

See supra notes 111-16 and accompanying text.

126

See, e.g., Email from Jessica Lin to George Cornell (Sept. 20, 2009, 18:28:38 EST)

(on file with author) (displaying an automatically generated voicemail transcript along with

the automatically generated time stamp and caller phone number).

127

Id.

128

Id.; see Pogue, supra note 3.

129

See United States v. Hamilton, 413 F.3d 1138, 1142-43 (10th Cir. 2005) (noting that

timestamps and screen name information accompanying pornograhic images were

computer-generated information that could not qualify as statements subject to exclusion as

hearsay).

130

State v. Armstead, 432 So. 2d 837, 839-40 (La. 1983) (“the printout of a telephone

trace in this type of system does not represent evidence of computer stored declarations. The

computer generated data by recording the source of various telephone connections as it was

making them . . . . [T]he evidence in this case was generated solely by the electrical and

mechanical operations of the computer and telephone equipment, and was not dependent

upon [human observations or reporting].”); State v. Carr-Poindexter, No. Civ.A. 20197,

2005 WL 737371, at *7 (Ohio App. 2 Dist. Apr. 1, 2005) ( “[C]aller ID information

provided to a telephone user is based on computer-generated information and not simply

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

computer-generated portions of an email conveying a Google Voice transcript

(i.e. the time, date, and caller number) can be distinguished from the

automatically transcribed statements. The former are non-hearsay

components,

131

while the latter, if offered to prove the truth of statements

contained therein, may be hearsay.

132

Therefore, a proponent that seeks to

admit such an email to prove a fact supported by its computer-generated

portions, such as the fact that a voicemail was received by a certain number at

a certain time, will be more likely to succeed if he or she can persuade the

judge to distinguish between the email’s computer-generated and computer-

stored components.

133

If the transcript portion of the email contains human

assertions that would be hearsay or otherwise improper to admit along with the

computer-generated information, the proponent could submit a redacted copy

or seek an instruction from the bench that limits the jury’s consideration to the

computer-generated, non-hearsay components.

134

5. Getting Around the Hearsay Exclusion: Exceptions to the Hearsay

Rule

Although the transcript portions of an email from Google Voice are

susceptible to classification as hearsay,

135

is there a reason why the exceptions

that allow admission of hearsay relayed through other mediums would not

apply to automatically transcribed voicemails? The hearsay exception rules

have already been applied to media that are closely analogous to automatic

voicemail transcriptions, such as recordings of telephone calls and transcripts

repetition of prior recorded human output or observation, and thus does not fall within the

scope of the hearsay rule. . . . Caller ID evidence, therefore, will not be inadmissible on

hearsay grounds, but may be attacked based on a lack of foundation regarding the reliability

of the device, or by otherwise demonstrating the unreliability of the information disclosed

by it.”) (citations omitted).

131

See supra notes 128-30 and accompanying text.

132

See supra notes 120-24 and accompanying text.

133

Wolfson, supra note 102, at 167-68 (outlining steps by which a judge can analyze

whether certain computer outputs are computer-stored information or computer-generated

non-hearsay and noting that “[t]his allows attorneys to break down the record into its

constituent parts, which are usually a combination of computer-stored and computer-

generated data.”).

134

See FED. R. EVID. 105 (“When evidence which is admissible as to one party or for one

purpose but not admissible as to another party or for another purpose is admitted, the court,

upon request, shall restrict the evidence to its proper scope and instruct the jury

accordingly.”).

135

See supra notes 68-108, 117-23 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

of such recordings. Why should courts not treat automatically generated

transcripts like other forms of recorded communication? In United States v.

Boyd, the 4th Circuit held that a district court did not abuse its discretion in

admitting a recording and a transcript of a 911 call by an eyewitness to a

shooting.

136

Though the recorded statement was clearly an out-of-court

statement, the court admitted the recording and transcript under the excited

utterance exception to the hearsay rule because the declarant had made the

statements contemporaneously with the shooting, while in a state of

excitement.

137

In United States v. Hawkins, the Eighth Circuit addressed

whether a recording of a 911 call could come within the present sense

impression exception to the ban on hearsay.

138

The court held that the caller’s

out-of-court statements from the 911 tape were admissible as a present sense

impression because the caller’s statements were contemporaneous with the

events being described.

139

Thus, the Federal Rules of Evidence “may provide

hearsay exceptions for electronically stored communications containing either

present sense impressions or excited utterances.”

140

The category of

electronically stored communications includes email, which may survive a

hearsay objection through the exception for “then existing mental, emotional,

or physical condition[s]” laid out in Federal Rule of Evidence 803(3).

141

136

See United States v. Boyd, 237 F. App’x 892, 893 (4th Cir. 2007).

137

See id. at 893 (citing FED. R. EVID. 803(2)) (“Melvin James’s statements during the

911 call were admissible under the excited utterance exception to the hearsay rule, which is

‘[a] statement relating to a startling event or condition made while the declarant was under

the stress of excitement caused by the event or condition.’”).

138

United States v. Hawkins, 59 F.3d 723, 730 (8th Cir. 1995), vacated on other

grounds, 516 U.S. 1168 (1996).

139

“[S]tatements from the 911 tape were admissible as a ‘present sense impression’

under Rule 803(1). Under that rule, a court may admit as an exception to the hearsay rule

‘[a] statement describing or explaining an event or condition made while the declarant was

perceiving the event or condition, or immediately thereafter,’ even though the declarant is

available to testify.” Id. (citing FED. R. EVID. 803(1)).

140

Lorraine v. Markel American Ins. Co., 241 F.R.D. 534, 569 (D. Md. 2007).

141

See id. at 570 (“Rule 803(3) is particularly useful when trying to admit e-mail, a

medium of communication that seems particularly prone to candid, perhaps too-candid,

statements of the declarant’s state of mind, feelings, emotions, and motives.”); FED. R. EVID.

803(3) (excluding from the hearsay ban “[a] statement of the declarant’s then existing state

of mind, emotion, sensation, or physical condition (such as intent, plan, motive, design,

mental feeling, pain, and bodily health), but not including a statement of memory or belief

to prove the fact remembered or believed unless it relates to the execution, revocation,

identification, or terms of declarant’s will.”).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

one district court observed, “Given the widely accepted fact that most writings

today are created and stored in electronic format, it is easy to see that the many

types of documents and writings covered in Rule 803 will implicate electronic

writings.”

142

If one views automatically transcribed voicemails as just another of the

several electronic mediums through which human statements are

memorialized, what would prevent the application of these same hearsay

exceptions to such transcripts? One characteristic that could set automatic

transcriptions apart from other mediums like email and audio recordings is the

level of inaccuracy with which the software converts human oral statements

into written form.

143

One might argue that a major rationale behind the limited

exceptions to hearsay is that the excepted types of statements are thought to be

more reliable than other, inadmissible forms of hearsay.

144

Rule 803’s

exceptions to hearsay “[proceed] upon the theory that under appropriate

circumstances a hearsay statement may possess circumstantial guarantees of

trustworthiness sufficient to justify nonproduction of the declarant in person at

the trial even though he may be available.”

145

One might argue that Google

Voice and other automatic transcription programs are too inaccurate to produce

trustworthy records of out-of-court statements. Therefore, the argument goes,

the rationale underlying the hearsay exceptions cannot apply to hearsay

statements contained in automatically generated transcripts.

The preceding argument, however, ignores the nature of the trustworthiness

on which the hearsay exceptions are predicated. “The underlying theory of

[the present sense impression exception codified in Federal Rule of Evidence

803(1)] is that substantial contemporaneity of event and statement negative the

likelihood of deliberate or conscious misrepresentation.”

146

“The theory of

142

Lorraine, 241 F.R.D. at 568.

143

See, e.g., Gallagher, supra notes 8-9 (showing several examples of highly inaccurate

Google Voice transcriptions under challenging circumstances); Email from Jessica Lin to

George Cornell, supra note 126 (in which the mock voicemail of the author mumbling “Uh,

you know for someone who’s supposed to be my attorney, you never seem to be available -

I’m a little bit ticked off about this and, uh, I’m really hoping that your law firm, uh,

Bing’em Bang’em and Bill’em has uh, shreaded those documents that I wanted shreaded

before. . . .” was transcribed to read “You know, for summers with you. My attorney you

never seem to be available. I’m a little bit ticked off about this and i [sic] really hoping that

your law firm thing Banking Bill has a dreaded those documents. I wanted to insure the best

for you . . . .”).

144

See FED. R. EVID. 803 advisory committee’s note.

145

Id.

146

FED. R. EVID. 803(1) advisory committee’s note (emphasis added).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

[Federal Rule of Evidence 803(2)’s exception for excited utterances] is simply

that circumstances may produce a condition of excitement which temporarily

stills the capacity of reflection and produces utterances free of conscious

fabrication.”

147

Federal Rule of Evidence 803(3)’s exception for then existing

mental, emotional, or physical condition “is essentially a specialized

application of [Federal Rule of Evidence 803(1)].”

148

The advisory committee

notes to Federal Rule of Evidence 803(4)’s exception for statements made for

the purposes of medical diagnosis or treatment state that “[e]ven those few

jurisdictions which have shied away from generally admitting statements of

present condition have allowed them if made to a physician for purposes of

diagnosis and treatment in view of the patient’s strong motivation to be

truthful.”

149

The committee’s preoccupation with dangers like fabrication,

misrepresentation, and motivations for truthfulness, suggest that the

trustworthiness it had in mind is largely, if not primarily, concerned with

truthfulness and sincerity.

150

“With a machine, however, there is no possibility

of a conscious misrepresentation, and the possibility of inaccurate or

misleading data only materializes if the machine is not functioning

properly.”

151

Thus, while statements conveyed by the transcript may be

inadmissible hearsay if they do not meet the conditions of any exception,

152

any statement that would otherwise meet an exception would not become

insincere or untruthful merely by virtue of its transcription by a computer.

153

In short, a computer program that performs automatic transcription may be

inaccurate,

154

but it cannot be a liar.

155

There is no greater risk of

untruthfulness or insincerity presented by these transcripts than by other forms

of electronically stored information to which the exceptions apply.

156

Therefore, hearsay exceptions should apply to automatically transcribed

voicemails to the same extent that they apply to analogous ways of preserving

147

FED. R. EVID. 803(2) advisory committee’s note (emphasis added).

148

FED. R. EVID. 803(3) advisory committee’s note (emphasis added).

149

FED. R. EVID. 803(4) advisory committee’s note (emphasis added).

150

See supra notes 146-49 and accompanying text.

151

State v. Armstead, 432 So. 2d 837, 840 (La. 1983).

152

See supra notes 68-108, 117-23 and accompanying text.

153

See Armstead, 432 So. 2d at 840 (noting that machines are incapable of “conscious

misrepresentation”).

154

See supra note 143 and accompanying text.

155

See Armstead, 432 So. 2d at 840 (noting that machines are incapable of “conscious

misrepresentation”).

156

See supra notes 136-42, 151 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

human statements, such as emails and audio recordings.

B. The Requirement of Authentication or Identification

1. Federal Rule of Evidence 901

In addition to the rule against hearsay, the authentication requirement might

serve to bar the admission of automatic voicemail transcriptions into

evidence.

157

For federal courts and courts in those states that have adopted the

Federal Rules of Evidence, the authentication requirement is codified in

Federal Rule of Evidence 901.

158

Federal Rule of Evidence 901 provides that

“[t]he requirement of authentication or identification as a condition precedent

to admissibility is satisfied by evidence sufficient to support a finding that the

matter in question is what its proponent claims.”

159

“This requirement of

showing authenticity or identity falls in the category of relevancy dependent

upon fulfillment of a condition of fact, and is governed by the procedure set

forth in Federal Rule of Evidence 104(b).”

160

In Huddleston v. United States,

the Supreme Court specified that in order to determine whether enough

evidence has been submitted to satisfy the conditional relevance requirement

of Federal Rule of Evidence 104(b), “the [trial] court simply examines all the

evidence in the case and decides whether the jury could reasonably find the

conditional fact . . .by a preponderance of the evidence.”

161

Therefore, in

determining whether to admit an item of evidence, a judge must be persuaded

that there is enough foundation laid so that a jury could reasonably find, by a

preponderance of the evidence, that the offered evidence is what the proponent

claims it to be.

162

Federal Rule of Evidence 902 lists several types of evidence

that are self-authenticating, but does not include anything resembling an

automatically transcribed voicemail.

163

Federal Rule of Evidence 901(b) lists ten methods by which a proponent of

157

See FED. R. EVID. 901(b) advisory committee’s note (“It should be observed that

compliance with requirements of authentication or identification by no means assures

admission of an item into evidence, as other bars, hearsay for example, may remain.”).

158

FED. R. EVID. 901.

159

FED. R. EVID. 901(a).

160

FED. R. EVID. 901(a) advisory committee’s note.

161

Huddleston v. United States, 485 U.S. 681, 690 (1988).

162

See supra notes 159-61 and accompanying text; FISHER, supra note 67, at 806 (“The

Supreme Court’s elaboration of Rule 104(b)’s standard in Huddleston . . . presumably

applies in [the Rule 901] context as well.”).

163

See FED. R. EVID. 902.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

evidence may establish its authenticity.

164

The Rule makes it clear that the

methods shown are “[b]y way of illustration only, and are not by way of

limitation.”

165

A proponent of evidence may therefore choose to use several of

the methods in tandem, or dispense with the listed options and find an

alternative way to authenticate the evidence.

166

The following discussion

focuses on several of these illustrations, and evaluates their feasibility as

methods to authenticate automatic voicemail transcriptions. One would expect

that most proponents of such transcripts would seek to prove that the transcript

in question is an accurate representation of what a certain caller said while

leaving a voicemail. If so, the proponent is concerned with showing both (1)

that the transcript accurately reflects the words spoken, and (2) that the words

spoken were those of the alleged caller. The accuracy problems with current

voicemail transcription software will complicate efforts to prove the first

point.

167

In addition, “[some] courts are concerned that the information

generated for use in litigation may have been altered, changed or manipulated

after its initial input, or that the programs and procedures used to create and

maintain the records are not reliable or accurate.”

168

On the other hand, the

standard of proof for authentication is relatively lenient when compared to

other standards.

169

Also, other courts apply concerns about tampering and

inaccuracy of electronic records to the weight of the evidence rather than its

authenticity.

170

2. Authentication by a Witness to the Conversation

The first illustrative way of establishing authenticity is to present testimony

164

See FED. R. EVID. 901(b).

165

Id.

166

FISHER, supra note 67, at 806.

167

See Gallagher supra notes 8-9, 143 and accompanying text.

168

Lorraine v. Markel Am. Ins. Co., 241 F.R.D. 534, 574 (D. Md. 2007) (discussed in

the context of the business records exception to hearsay).

169

See FED. R. EVID. 901(a) and advisory committee’s notes;

Huddleston v. United States, 485 U.S. 681, 690 (1988); FISHER, supra note 67, at 806 (judge

need only be satisfied that jury could find authenticity by a preponderance of the evidence)

(emphasis added).

170

See United States v. Safavian, 435 F. Supp. 2d 36, 40-41 (D.D.C. 2006) (holding that

emails at issue, which the defendant argued were untrustworthy because they had been

forwarded through other emails in the process of which they could have been modified,

were sufficiently authenticated and that “[t]he defendant’s argument is more appropriately

directed to the weight the jury should give the evidence, not to its authenticity.”).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

of a witness with knowledge, or “[t]estimony that a matter is what it is claimed

to be.”

171

This method “contemplates a broad spectrum ranging from

testimony of a witness who was present at the signing of a document to

testimony establishing narcotics as taken from an accused and accounting for

custody through the period until trial . . ..”

172

Courts have regarded transcripts

as properly authenticated when a participant in the transcribed conversation

gives unrebutted testimony that the transcripts are correct to the best of the

participant’s knowledge.

173

Therefore, unrebutted testimony from the caller

who left the voicemail from which a transcript is automatically generated could

lay an adequate foundation for authenticating such transcripts.

174

3. Authentication where Witnesses to the Conversation Are

Unavailable: Identifying the Caller

What options would a proponent have in the likely scenario in which the

transcribed statements are against the interest of the caller, and the only human

participant in the conversation is either unwilling or unavailable to testify as to

their authenticity? Evidence of “telephone conversation[s] is admissible

provided that the identity of the speaker is satisfactorily established.”

175

“The

cases are in agreement that a mere assertion of one’s identity by a person

talking on the telephone is not sufficient evidence of the authenticity of the

conversation and that additional evidence of his identity is required.”

176

Hence, Federal Rule of Evidence 901(b)(6), which provides an illustration of

authenticating a telephone conversation, involves the identification of the

person called rather than the caller.

177

In the example, the person called is

171

FED. R. EVID. 901(b)(1).

172

FED. R. EVID. 901(b)(1) advisory committee’s note.

173

See United States v. Wright, 932 F.2d 868, 880 (10th Cir. 1991) (holding that “the

unrebutted testimony of one of the participants in the actual conversations is sufficient

authentication” of transcripts of a recorded conversation).

174

See id.

175

United States v. Biondo, 483 F.2d 635, 644 (8th Cir. 1973).

176

FED. R. EVID. 901(b)(6) advisory committee’s note.

177

See FED. R. EVID. 901(b)(6). A rationale for this distinction is that “[t]he number

called is owned by and under the control of the person to whose name the number is

attached. It is certain that the answer is from that number, a circumstance tending to show

that the person answering is the person called or one who has authority to answer for him.

But where the telephone call is from an unknown number and the person called answers and

asks who it is, any reply as to the number from which the call comes or as to the name of the

caller would be pure hearsay. There would be no competent evidence that the call came

from the number it claimed to be.” State ex rel. Strohfeld v. Cox, 30 S.W.2d 462, 464 (Mo.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

identified through a combination of (1) evidence establishing that the phone

number dialed is associated with the person allegedly called, and (2) additional

circumstances such as the self-identification by the person called.

178

Thus, it

seems unlikely that a proponent would be able to adequately link an

automatically transcribed voicemail to an alleged caller by relying solely on

statements of self-identification contained in the transcript.

179

The advisory

committee notes to Federal Rule of Evidence 901(b)(6) suggest that alternative

ways to establish the identity of the caller include (1) evidence that the

recipient of the call recognized the caller’s voice, or (2) evidence of other

special characteristics of the content of the call that link it to the caller.

180

Recognition of the voice that made the call is clearly inapplicable in this case,

since the proponent seeks to admit a transcript of the voicemail rather than a

recording of the audio. However, one may establish the identity of a party to a

telephone conversation by circumstantial evidence.

181

Federal Rule of

Evidence 901(b)(4) states that authenticity may be established through

evidence of distinctive characteristics such as “[a]ppearance, contents,

substance, internal patterns, or other distinctive characteristics, taken in

conjunction with the circumstances.”

182

“Thus a document or telephone

conversation may be shown to have emanated from a particular person by

virtue of its disclosing knowledge of facts known peculiarly to him.”

183

addition, language patterns may indicate authenticity.

184

Accents are not

reflected in Google Voice transcripts, except to the extent that they seem to

cause errors in many cases.

185

However, certain grammatical errors and other

patterns in the words that a caller has a habit of using may very well carry over

into the transcript.

186

As for relaying facts known peculiarly to the caller, the

1930).

178

See FED. R. EVID. 901(b)(6).

179

See supra notes 175-78 and accompanying text.

180

See FED. R. EVID. 901(b)(6) advisory committee’s note (“The additional evidence

need not fall in any set pattern. Thus the content of his statements or the reply technique . . .

or voice identification . . . may furnish the necessary foundation.” (citing Fed. R. Evid.

901(b)(4)-(5)).

181

United States v. Espinoza, 641 F.2d 153, 170 (4th Cir. 1981).

182

FED. R. EVID. 901(b)(4).

183

FED. R. EVID. 901(b)(4) advisory committee’s note.

184

Id.

185

See, e.g., Gallagher, supra note 9 (“A number of people tried to stump Google with

their wacky accents. As you can see, the results were mixed.”).

186

See, e.g., id. (several of the examples posted accurately reflected contractions in

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

fact that computers do not consciously fabricate or mislead provides assurance

that this method of proving authenticity is appropriate for automatically

generated transcripts in the same way that it is appropriate for letters and

telephone calls.

187

In other words, there is little risk that the computer will

spontaneously fabricate facts that would only be known to a person other than

the actual caller.

188

Therefore, it does not seem that the risk of false positive

authentications is significantly increased by allowing the use of this method.

189

Another source of circumstantial evidence which a proponent could use to

tie an alleged caller to a Google Voice transcript is the telephone number of the

caller. The telephone number of the telephone which was used to record the

voicemail is automatically generated and included with the voicemail transcript

that a Google Voice customer receives.

190

Courts have held that the automatic

output of caller identification information (such as from caller ID devices) is a

corroborative circumstance that can be used to identify the caller in telephone

conversations for authentication purposes.

191

However, the proponent of this

information has the burden of proving that the device displaying the caller’s

information is reliable.

192

4. Authentication where Witnesses to the Conversation Are

Unavailable: Authenticating the Substance of the Transcript

Assuming that the above methods are sufficient to tie an automatically

generated transcript to a certain caller, how can a proponent of one of these

transcripts lay enough foundation to show that it is a written version of what

the caller said? When a transcript is offered, the authentication question is

usually whether the transcript is an accurate rendition of a recorded

conversation.

193

Usually, transcripts can be authenticated either through

testimony by a party to the conversation

194

or the testimony of the

transcriber.

195

If the only party to the conversation is unwilling or unavailable

words).

187

See supra note 151 and accompanying text.

188

See id.

189

See supra notes 185-87 and accompanying text.

190

See supra notes 127-30 and accompanying text.

191

People v. Caffey, 792 N.E.2d 1163, 1191 (Ill. 2001).

192

Id.

193

United States v. Devous, 764 F.2d 1349, 1355 (10th Cir. 1985).

194

United States v. Wright, 932 F.2d 868, 880 (10th Cir. 1991).

195

Devous, 764 F.2d at 1355 (citing United States v. Rochan, 563 F.2d 1246, 1251 (5th

Cir. 1977)).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

to authenticate the transcript, what other authentication options are available

when the transcriber is a computer program? Federal Rule of Evidence

901(b)(9) indicates that a proponent can authenticate the output of a process or

system by presenting “[e]vidence describing a process or system used to

produce a result and showing that the process or system produces an accurate

result.”

196

While the rule is phrased as if only the process itself is scrutinized,

in practice an adequate foundation also requires specific proof indicating that

the underlying method was properly followed or put into execution in the

particular case.

197

Usually evidence produced by a process or system is

presented to the trier of fact by means of expert testimony, and the same expert

lays the foundation as to the accuracy of the process and the methods followed

in its execution.

198

However, “where the process or system is new or

controversial, there exists a possibility that even a personal endorsement by an

expert witness will not suffice to make the evidence admissible . . ..”

199

For

computer outputs, authentication under Federal Rule of Evidence 901(b)(9)

can be accomplished by evidence that:

(1) the computer equipment is accepted in the field as standard and

competent and was in good working order; (2) qualified computer

operators were employed; (3) proper procedures were followed in

connection with the input and output of information; (4) a reliable

software program was utilized; (5) the equipment was programmed and

operated correctly; and (6) the exhibit is properly identified as the output

in question.

200

Some of these factors seem difficult to apply to the process of automatic

voicemail transcription and seem tailored more for scientific or business fields.

How does one define, for example, the proper procedure for leaving a

voicemail? If expert testimony is required on these subjects, it seems likely

that only someone familiar with Google Voice’s transcription program would

be qualified to testify about the nature of the process and its accuracy.

However, Google’s own statements concede that there are reliability issues

with the software.

201

In fact, the program itself indicates that it is unsure about

196

FED. R. EVID. 901(b)(9).

197

4 MUELLER & KIRKPATRICK, supra note 98, at § 9:20.

198

Id.

199

Id.

200

Id.

201

See About voicemail transcriptions - Google Voice Help, supra note 7 (“This is the

only fully automated voicemail transcription on the market. This means, however, that it’s

not perfect yet. It will improve over time as our transcription engine gets smarter. The

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

certain transcribed words by displaying them in a lighter shade.

202

Taken

together with examples of the inaccuracy of the program,

203

it seems that a

party seeking to introduce such a transcript would have a difficult time

convincing a jury by preponderance of the evidence that the transcription

program produces reliable results. Therefore, without an actual witness to the

conversation to lay the foundation as to a transcript’s accuracy, it will be

difficult to find an alternative means of authenticating automatically

transcribed voicemails.

C. The Rule 403 Balancing Test: the Polygraph Analogy

Though neither the rule against hearsay nor the authentication requirements

of the Federal Rules of Evidence seem to serve as a per se bar against

admission of automatic voicemail transcriptions, could courts nonetheless

decide that such evidence is per se inadmissible? If yes, it would not be the

first time that evidence based largely on a machine’s output was treated as

such. In the landmark 1923 case of Frye v. United States, the D.C. Court of

Appeals held that the systolic blood pressure deception test (a precursor to the

modern polygraph or lie detector test) had not yet gained enough acceptance in

the scientific community to permit admission of evidence based on such

tests.

204

For the seven decades before Daubert v. Merrell Dow

Pharmaceuticals specified a new test for the admissibility of scientific

evidence,

205

the Frye decision effectively served as a per se bar against

admission of polygraph evidence.

206

Because Daubert and Frye are primarily

concerned with the reliability of expert testimony based on interpretation of

scientific data,

207

they seem inapplicable to automatically transcribed

voicemails, which do not involve a human interpreter.

208

However, some

courts have continued to maintain Frye’s presumptive ban on polygraph

quality of the transcripts will vary depending on the caller, the background noise, and

whether the caller is using a microphone.”).

202

Pogue, supra note 3.

203

See Gallagher, supra notes 8-9, 143 and accompanying text.

204

Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923).

205

See Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 592-96 (1993) (outlining a

“flexible” approach to judging the admissibility of scientific testimony that looks to more

than just whether the methods involved are generally accepted in the scientific community).

206

John E. Theuman, Annotation, Admissibility in Federal Criminal Case of Results of

Polygraph (Lie Detector) Test - Post - Daubert Cases, 140 A.L.R. FED. 525, §2[a] (1997).

207

See id.

208

See Pogue, supra note 3 (discussing Google Voice’s fully automated nature).

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

evidence through a stringent application of Federal Rule of Evidence 403.

209

In a similar manner, Rule 403 might be used to justify broad prohibitions

against the admission of automatic voicemail transcriptions. Rule 403

provides that, “[a]lthough relevant, evidence may be excluded if its probative

value is substantially outweighed by the danger of unfair prejudice, confusion

of the issues, or misleading the jury . . ..”

210

In the polygraph context, courts

that exclude such evidence point to the technology’s limited probative value.

211

To the extent that unreliable evidence lacks probative value, other courts

phrase their rejection of polygraph evidence in terms of its unreliability.

212

the same time, courts are concerned with the substantial prejudicial effects that

polygraph evidence can produce. For example, there is a fear that jurors will

treat the results of polygraph tests as conclusive proof of a witness’s honesty,

thus allowing technology to usurp the jury’s role in assessing credibility and

guilt.

213

Polygraph evidence may also be highly prejudicial because it (1) has

an “aura of near infallibility,” (2) often produces “an opinion regarding the

ultimate issue in the case,” and (3) determining its reliability consumes a great

deal of judicial resources.

214

Could litigants use a similar Rule 403 analysis to bar admission of automatic

voicemail transcriptions? Although such transcripts do not involve an expert

human intermediary, like in the case of polygraphs, it does seem that some of

the same concerns apply. Looking to the probative value of automatically

transcribed voicemails, the rampant inaccuracies of the transcription software

would serve to reduce the probative value of such transcripts.

215

One could

also make the case that admitting such transcripts comes with a high risk of

209

See, e.g., United States v. Sherlin, 67 F.3d 1208, 1216-17 (6th Cir. 1995) (holding

that, despite the new Daubert standard for analyzing the admissibility of scientific expert

testimony, a general rule against admission of polygraph testimony could endure because,

under FED. R. EVID. 403, the probative value of such evidence is often outweighed by the

risk of unfair prejudice).

210

FED. R. EVID. 403.

211

See Sherlin, 67 F.3d at 1216-17 (holing that the probative value of polygraph

evidence is particularly “dubious” when commissioned unilaterally by a defendant).

212

See United States v. Lea, 249 F.3d 632, 639-40 (7th Cir. 2001) (noting that, whether

admissibility is analyzed under Daubert or FED. R. EVID 403, “[a]s the reliability of the

evidence decreases, the likelihood increases that the probative value may be substantially

outweighed by the prejudicial effect.”).

213

See id. at 639.

214

United States v. Crumby, 895 F. Supp. 1354, 1356 (D. Ariz. 1995) (admitting

polygraph evidence).

215

See supra notes 8-9, 143 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

unfair prejudice. In the same way that jurors might erroneously view an expert

in polygraph testing as nearly infallible,

216

they may regard a computer

program with the same unquestioning deference. In the case of polygraphs,

proponents of lie detector evidence can reduce the risk of juror deference by

addressing the limits of the technology themselves.

217

Proponents of

automatically transcribed voicemail evidence could use a similar approach to

dispel concerns that the jury will blindly accept a transcription program’s

output. Likewise, any prudent litigant seeking to lessen the impact of such

transcripts can also be expected to highlight the technology’s flaws.

Many of the particular concerns that may justify exclusion of polygraph

evidence are inapplicable to automatically transcribed voicemails. For one, it

is hard to see how automatically transcribed voicemails usurp the role of the

jury since, unlike polygraphs,

218

transcripts do not purport to determine

sincerity or guilt. Second, while such transcripts may bear on a key fact or

issue in a case, it would be too much of a stretch to call them “an opinion

regarding the ultimate issue in the case . . ..”

219

If an admissible opinion is

“helpful to a clear understanding of the witnesses’ testimony or the

determination of a fact in issue,”

220

then it is hard to see how a transcription

program aides the jury’s understanding any more than a recording of the

voicemail. Finally, while determining the reliability of automatic voicemail

transcription programs on a case-by-case basis will undoubtedly consume

some judicial resources, the alternative of placing a general bar on such

evidence is shortsighted. Because the algorithms that convert voicemail

messages into transcripts are constantly improving,

221

such a per se bar would

risk excluding more reliable transcriptions in the future. In fact, the

improvement in lie detection technology is one reason why some courts have

departed from the old per se rule against polygraph evidence.

222

216

See Crumby, 895 F. Supp. at 1356.

217

See Theuman, supra note 206, at § 2[b] (suggesting that “it may be advisable for

counsel to caution witnesses not to make any exaggerated claims regarding the technique’s

capabilities or reliability, but rather to be candid about the technique’s limitations.”).

218

See Lea, 249 F.3d at 639 (discussing how polygraph testing usurps the jury’s role of

assessing credibility and guilt).

219

Crumby, 895 F. Supp. at 1356 (emphasis added).

220

FED. R. EVID. 701.

221

See About voicemail transcriptions - Google Voice Help, supra note 7

(acknowledging Google Voice’s flaws but promising that it “will improve over time as our

transcription engine gets smarter.”).

222

See, e.g., Crumby, 895 F. Supp. at 1357-58 (listing as one of its main reasons for

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

B.U. J. SCI. & TECH. L. [Vol. 17

Thus it seems difficult to justify a per se rule against admitting

automatically transcribed voicemails based on Rule 403’s balancing test.

223

This is not to say that proponents of such transcript evidence will always

succeed in getting past Rule 403. For instance, the reliability of the

transcriptions might be particularly bad in certain circumstances,

224

rendering

their probative value particularly low. If the transcribed voicemail at issue also

discussed the caller’s irrelevant deviant exploits, it could be a clear case of

unfair prejudice substantially outweighing probative value.

IV. CONCLUSION

The Federal Rules of Civil Procedures’ inclusive definition of electronically

stored data is broad enough to encompass emails containing automatically

transcribed voicemails. Past court experience with the rise of email and

voicemail as discoverable information suggests that it is only a matter of time

before these transcripts also become part of the ever-growing amount of

electronically stored information involved in litigation.

The value of automatically transcribed voicemail messages to litigants will

depend in large part on their admissibility. I have argued that the rule against

hearsay is not an absolute bar to admission of these transcripts. Instead,

admissibility will turn on whether the proponent of such evidence seeks to

introduce information from the computer-generated or the computer-stored

components of transcription results. If the proponent seeks to introduce

computer-stored, human statements from the transcript for the truth of the

matters asserted therein, he or she is vulnerable to hearsay objections.

However, an analysis of the Federal Rules of Evidence exceptions to the

hearsay ban indicates that several of these exceptions would also apply to

qualifying transcripts.

Authentication poses a challenge to any proponent seeking to admit

automatically transcribed voicemail evidence. While the Federal Rules of

Evidence and case law interpreting them specify a fairly lenient burden of

proof for authentication, the more rigorous authentications required for

computer outputs and transcripts may serve to preclude automatically

transcribed voicemails. This is likely to be more of a problem, however, when

no witness to the original telephone call is available to authenticate the

transcript.

admitting polygraph evidence “a significant increase in the reliability of polygraph evidence

over recent years.”).

223

See supra notes 210-22 and accompanying text.

224

See supra notes 8-9 and accompanying text.

THIS VERSION DOES NOT CONTAIN PARAGRAPH/PAGE REFERENCES.

PLEASE CONSULT THE PRINT OR ONLINE DATABASE VERSIONS FOR

PROPER CITATION INFORMATION.

2011] AUTOMATICALLY TRANSCRIBED VOICEMAIL MESSAGES

Federal Rule of Evidence 403 should not serve as a per se bar to the

admission of automatically transcribed voicemails. While there will likely be

cases where a transcript’s probative value is substantially outweighed by its

risk of unfair prejudice, an absolute bar against this entire class of evidence is

unwarranted. Many of the policy concerns that courts have used to justify the

retention of a per se exclusion of polygraph evidence do not apply to

automatically transcribed voicemails. Furthermore, either party in a litigation

can mitigate the risk that the jury will place undue faith in the accuracy of

these transcription programs by identifying their shortcomings.

Because of these challenges to admissibility, the evidentiary value of

automatically transcribed voicemails will vary dramatically with the facts of

each case, but they will not be per se excluded as a viable class of evidence.