The Discourse of Voicemail

This paper attempts to determine to what degree voicemail messages can be considered a discourse genre – that is, to what degree and in what ways they appear to be uniform across speakers. Thirty-seven voice messages were recorded from the cellular phones of three University of Michigan students. The messages were analyzed in terms of their overall structure, the discursive functions that were executed therein, and the speciﬁ c words, phrases and prosodic strategies that were used to execute certain functions. The messages were found to have highly uniform openings and closings, and the message bodies were found to reduce to a small set of discursive functions. In addition, certain words, phrases and devices appeared frequently and in predictable locations within the messages. It is concluded that voicemail message-leaving is a highly structured act governed by conventions that arise both from face-to-face conversation and from the speciﬁ c constraints of the medium.


Introduction
In an age in which technology increasingly facilitates long-range communication, voicemail has become a ubiquitous feature of American life. Voicemail systems of one kind or another are widespread in both domestic and institutional contexts, and the ability to leave an effective and appropriate message is an essential part of many people's lives.
Answering machines, the predecessors of modern digital voicemail 1 , were originally used only by businesses or working professionals (Gold 1991: 243). Their usage spread, however, and by the mid 80's the an-ing and hanging up the minute they realized they were talking to a recording." Abby's suggestion: "Hire an answering service which performs the same duties as a private secretary. People who call you," she advised, "want to speak to a real person, not to a machine" (quoted in Dubin 1987: 29). In fact, declining to leave a message seems to be the preferred strategy wherever answering machines and voicemail are relatively uncommon (Dingwall 1992).
With enough exposure, however, people overcome their inhibitions and learn to negotiate the constraints imposed by a particular communicative medium. Any "recurrent communicative problem" will give rise in time to what Luckmann (cited in Alvarez-Caccamo/Knoblauch 1992: 476) 3 calls "communicative genres": "pre-cut communicative patterns which provide solutions" to these problems (Alvarez-Caccamo/ Knoblauch 1992: 476). This study aims to determine to what extent and in what ways voicemail has become such a genre.
Previous researchers have found in voice messages a three-part structure: an opening, a body, and a closing section (with one researcher including an optional 'postscript' section) (Alvarez-Caccamo/Knoblauch 1992; Liddicoat, 1994;Gold, 1991). In addition, researchers have found that this structure remains constant across cultures (Alvarez-Caccamo/ Knoblauch 1992;Goutsos 2001;Liddicoat 1994).
This study adheres to this intuitive three-part model of voice messages, but it aims to take a closer look at the internal structure of each section. It describes the organizational structure of voice messages and seeks to answer the following questions: What discursive functions are executed, and which of those functions are most common? What specific phrases and strategies are used to execute each of these functions? To what degree are these phrases and strategies individual, and to what degree do they appear to be common to a wider culture? What discursive and prosodic strategies do people use to overcome the specifi c communicative challenges posed by voicemail? How much do voice messages follow a pre-set structure?

The data
Thirty-seven voice messages were recorded from the cellular phones of three University of Michigan students. The subjects are all between the ages of 18 and 20. All thirty-seven messages were received naturally; i.e., none of them was solicited for the purpose of this study. The messages were transcribed and for each message the subjects were asked to describe their relationship with the callee (close friend, parent, boss, co-worker, etc.) . The data comprises twenty-three speakers, several of whom left multiple messages. The data can be grouped as follows: 25 messages from 13 friends (all within two years of age of the subjects.) 5 messages from 4 acquaintances (all within two years of age of the subjects) 5 messages from 2 family members 3 messages from 3 institutions (2 from callee's place of work and 1 from a scholarship organization)

Transcription Conventions
Commas indicate short pauses; ellipses indicate longer ones. Periods mark the 'ends of sentences,' as defi ned primarily by 'prosody that creates a sense of conclusion' rather than by syntactic or semantic measures. XX indicates parts of the message that are indiscernible, while parentheses indicate speculative transcription of parts that are partially inaudible. An underscore marks a word that is phonetically lengthened (so_). A dash following a word marks a repair (messages [21] and [31] in Appendix). Names of persons are given as capital letters; a given letter or pair of letters represents the same person across all messages. Phone numbers are omitted, and the names of organizations and specifi c scholarships have been changed. (See Appendix for full transcriptions.) [7] Hey L it's O, Friend 5.

The Message Opening
[8] Hello, L, my love Friend 6.
[ There are three basic elements that typically appear in the message opening, in the following order: (1) a greeting word such as hi, hey or hello; (2) direct address to the callee, and (3) self-identifi cation. The message typically opens with a greeting word. The most common one in this data set is hey. The three institutional messages do not begin with hey, which, along with this author's intuition as a native speaker, suggests that it is a word best reserved for interaction with peers and family members. Hi and hello were the only other greeting words that appeared, though in far lower frequency than hey.
There are only fi ve messages in which none of these three opening words is used. Of these, four include some form of direct address, whether by name or by a moniker (e.g. bitch). In fact, 32 of the messages contain some sort of direct address, the most common form being the callee's name.
Finally, 30 out of 37 of the messages include some form of self-identifi cation. Again, the most common method is for the caller to state his or her name (and, in the case of institutional messages, the organization on whose behalf he or she is calling). A few, however, opt for it's me in place of their name.
There was only one message that lacked a greeting or address altogether [36]. In all, 27 of the messages contain openings of the following form: The presence of all these opening elements is curious for two reasons. First, a greeting (i.e. greeting word and name) usually constitutes part of an adjacency pair (the second part being a response greeting). Since the callee is absent, however, the adjacency pair must go uncompleted. Secondly, self-identifi cation seems to be redundant since most of the callers are readily identifi able to the callees by voice alone. (This fact was verifi ed by the polled subjects.) In light of these peculiarities, what functions might these opening elements be serving?
It has been suggested that the communicative procedures for leaving a voice message have been simply adapted from existing techniques in telephone-based communication (Liddicoat 1994). It is possible that this typical opening sequence is a derivative from normal telephone openings and that its form has been preserved incidentally. Alvarez-Caccamo and Knoblauch (1992: 496) have also convincingly demonstrated that voice messages, "instead of representing a disjunction in the verbal interactional history of any given pair of participants...are very productively embedded in such a history of callers and callees." It seems that callers treat their messages as interactional, despite the asynchronicity between the caller and the callee. Perhaps the adjacency pair formed by the greeting and caller address is intended to be completed at some later date, or perhaps the caller is completing a pair that was initiated in previous interaction with the callee. Gold (1991: 247) has also pointed out that "The bigger the mismatches in time and space between communicators, the more information that must be given in order to try and transcend the potential for miscommunication." In real-time conversation with friends and family members, self-identifi cation is inappropriate because it constitutes information that is presumed to be shared by both participants. It may be, however, that the need to avoid potential miscommunication via voicemail outweighs the pragmatic norms for when shared information needs to be made explicit (Gold 1991: 247).
I suggest that voicemail greetings also serve another practical function. In face-to-face interaction, participants are generally aware of each other's identity before conversation begins. With voicemail, callers' identities have to be established from their voices alone. If the caller were to launch into the body of the message without preamble, they would be burdening the message with a double task: making the identity of the caller known as well as carrying out whatever phatic or informational function it was designed to execute. The callee would be more likely to spend time processing who the message came from rather than apprehending its meaning. These formulaic opening sequences serve to prepare the callee, to give them time to set up in their minds a context for the message so that they can devote their cognitive resources to the body of the message.
Two of the openings depart from the normal schema in a noticeable manner: with something like a joke rather than a greeting word. That these opening were intended to be humorous was evident not only from the altered tone of voice in which they were delivered, but from the fact that they concluded with laughter on the part of the messageleaver before normal tone was assumed for the body of the message. Though these messages differ in form from the majority, however, they in fact seem to still serve the same functions. First, they prepare the callee for the message by delaying the start of the body of the message. And, since both messages come from people who are close friends of the callees, their voices alone may serve as identifi cation. (Note that caller [6] still identifi es himself after his joke: That's right bitch, it's me.) Both callers still address the callee by name, showing that they are still in some sense adhering to the expected opening routine. Both also exhibit interactional/conversational language: they both involve questions posed to the callees, and the second one even contains, in the question Did you just tell me to apres trois?, a reference to the message prompt 4 . In fact, part of the humor may derive from the fact that the callers deliberately take the interactive element too far in a medium that does not afford synchronous interaction. Most questions in the data are phrased indirectly (I was wondering if...); the fact that these direct questions cannot be answered in real time is part of what makes these openings funny. They may also be funny because the callers are aware that they are breaking established procedure in their openings. Despite this break, however, these openings serve the same discursive functions as more typical openings. Gold (1991: 252) has suggested that humor in voice messages may be a means of "mitigating discomfort (or irony)" arising from the peculiarity of the medium (parentheses hers). Again, such a sense of irony was probably much more common fi fteen years ago when voicemail was not so widespread, but it may be that there is still a lingering awareness of the mediating machine and that this awareness manifests as humor. Unfortunately, however, this study has no way of determin-ing whether humor occurs more frequently in voice messages than in everyday conversation.

Message Structures and Functions
The following table contains a break-down of the discursive functions executed in the body of the voice messages and the order in which they appear. Discourse markers that mark boundaries between these functions are given in brackets at the location they occur at in the message. Because words like so may have not only a discursive, but a lexical function, discourse markers were only transcribed as such when they were accompanied by pauses or distinctive prosody that clearly marked unit boundaries.  ...as well as one reference to a call that never came: References to past calls (real or expected) are generally encoded into the justifi cation for calling. There are also references to calls that the caller intends to make: These references to bothpast and anticipated communication show that the voice messages are grounded in a shared communicative history between the caller and callee. Of course, the majority of the messages come from family members and friends of the subjects, so this history is presumed; but the explicit references to this history make clear the prominence of voicemail in the subjects' daily lives. Voicemail is not just a practical tool, as it once was, but an important medium for everyday social interaction.

The Structure of the body
Justifi cation for calling tends to occur immediately following the opening, while well-wishing and requests for return calls tend to closely precede the closing. Outside of these patterns, however, there does not seem to be a fi xed order to the execution of different discursive functions within the message. It is likely that the presence or absence of the above categories and their ordering varies according to external factors like individual preference, how recently the caller and callee have spoken (if it all), etc. It is unfortunately not within the scope of this study to explore these factors; however, the basic functions identifi ed above could be used, it seems, in such studies in the future.

Boundaries between discourse functions
Boundaries between functional units are frequently marked with so, um and uh. Other markers like alright are common but are generally incorporated into sections rather than used to denote boundaries. (Alright, for example, frequently occurs in closings.) The extensive presence of discourse markers is another indication of the interactional origins and basis of voice messages. Callers seem in some ways to be creating one-sided dialogues (Gold 1991;Alvarez-Caccamo/Knoblauch 1992) or perhaps adapting techniques from normal real-time telephone conversation (Liddicoat, 1994).
Occasionally, sections are separated by a pause, but this seems to be a dispreferred marker. This may be related to a broader discomfort among Americans with long pauses in interactions; if so, then the separation of discursive sections by pauses is probably culturally variable. It is important to note, however, that because voicemail is an audioonly format, long pauses could be misconstrued by the hearer as signaling the end of the message, or some diffi culty on the part of the caller. They may as such be ruled out by the pragmatic requirements of the medium.

Humor in messages
11 of 37 of the messages incorporate some kind of joke or humorous device. That a given section is intended to be humorous is often made clear by the content, but it is always marked by distinct voicing or prosody; and it is bounded by laughter, a discourse marker, and fi nally a return to normal prosody. Two of the messages, [11] and [17], are delivered entirely in an altered voice that is clearly intended to be humorous. (This fact was also verifi ed with the two people who left the messages.) Though humor may be partially a response to irony, as indicated earlier, it may also be a response to the message prompt itself. The prompts of two of the subjects are humorous, and it was to these two subjects that 8 of the 11 humor-containing messages were sent. In fact, as seen earlier, one of the messages [35] contains a specifi c reference to the prompt: Research by Buzannell et al. (1996) suggests that students leaving messages on their professors' machines may choose their message content or style of delivery based on the prompt. Although in that study the callers were attempting to contact authority fi gures (so that they perhaps had a stronger motive to respond to what those authority fi gures said in their prompts), it is possible that callers in general are more likely to leave humorous messages when the prompt itself is humorous.
The choice to include humor is probably also personal. Five of the messages come from two speakers (1. and 22.), who included humor in each of their messages.

Message Closings
Speaker # Message # Message Text Relationship 1.

[1]
So, I will talk to you soon, I hope. OK. Bye.

Best friend
[2] so um uh, yeah just gimme a call when you get this. Alright? bye.

[3]
So um, just gimme a call whenever you are free, and, I love you and break a leg! Bye.

[4]
and hope all goes well and I will talk to you soon. Thanks bye. Friend 3.
[5] So talk to you later. Bye.

Friend [6]
but yeah gimme a call, and I will talk to you soon. Bye. 4.
[9] let me know and I'll see you later. Bye. Friend 7.
[10] I hope to hear from ya. Bye. Acquaintance
[12] I'm in the tech booth, come visit me. I love you. Bye.

[13]
Uh gimme a call, and we can go have dinner later too. OK, bye.

[14]
so um...just gimme a call and have a good class, and all that stuff. Love you. Bye.

[15]
Um, so yeah. Gimme a call when you get this. Love you, bye. 10.
[16] So, we'll see you a little bit after six, we're running late to rehearsal. Bye_.

[18]
I will talk to later and, yeah. Love you bye. 11.
[20] So I'll see you, uh, in a couple hours. Aright? See ya. Bye.

[21]
Ok? I love you, and I'll talk to you soon. bye.

[22]
Anyway, I love you, I'll talk to you soon. Bye.

[23]
and uh, and uh I love you and I'll talk to you soon. Bye.

[24]
and please let us know how it goes. I love you. Bye bye. Mother

[26]
If you can or cannot could you please gimme a call, the number is -------. Thanks.

16.
[27] call me back, let's hang out cause hopefully you're not so busy anymore. Alright, bye. Friend
[ In conversation, closings can consist of negotiation, verifi cation of the topics discussed, and/or simple goodbyes. Voicemail presents a unique problem for closing, however, since no interactional feedback is available. The singularity of this task was evident in the early days of answering machines, when people showed "a tendency towards inability to close" (Dubin 1987: 30). Dubin found that, except when making requests, people tended to simply fade out at the end of messages rather than constructing a defi nitive conclusion. Even in their more recent study, Alvarez-Caccamo and Knoblauch (1992: 494) argued that "there is an indefi nite variety of elements in the closing section [of voice messages], and messages can take very different forms." That variability seems not to be present in this data, however, which presents a remarkable uniformity in the closing sections. To begin with, 31 of the messages end in bye, and two end in bye bye. Of the remaining four, one contains bye, though it is not the fi nal word [11], and the other three conclude with thanks [26], [31], [34]. As with the opening sections, bye might seem to be a redundant feature of voice messages. After all, other signals -the sound of the caller hanging up, the machine's acknowledgment that the message is over -tell the callee when the message has concluded. Letters, emails, and other mediated genres do not display this feature, so why should voice mail? Again, it seems likely that saying farewell is a feature that has persisted from real-time conversation.
Words like bye also may indicate that there has been prior communication between the caller and the callee and suggest that there will be further communication between them (especially telephone communi-cation). This is supported by the observation that of the four messages that contain no word of farewell whatsoever, only one is from a person the callee described as a 'friend.' Of the other three, two are from institutions and one is from an acquaintance. Words of farewell, along with other features of spoken conversation, signal that the messages are grounded in a shared history between the caller and callee. Turns are signaled just as they are in spoken conversation, though the time lapse between responses is much greater.
In 20 out of 37 messages, the closing is immediately preceded by a request for a response on the part of the callee. Usually this is a request for a call back (call request), but in a few cases it is restricted to the more ambiguous let me know (response request). Here are some typical examples: Example 9 so um uh, yeah just gimme a call when you get this. Aright? Bye. In another 4 cases, a call-back request occurs very close to the closing, but with some other functional unit, such as an invitation, intervening.

Example 10
Uh gimme a call, and we can go have dinner later too. OK,bye. [13] Here there is a sort of invitation that separates the call-back request from the closing words. (Note: In [24], I love you is considered to be part of the closing, so it is included with the former category in Example 9. In [13], dinner has not been mentioned before, so the (indirect) invitation to dinner is considered to be executing a distinct function.) 5 of the messages contain some kind of well-wishing just before the closing: Example 11 hope all goes well and I will talk to you soon. The other 8 messages all contain some variation of I'll talk to you later or I'll talk to you soon immediately before the farewell. This is further evidence of the interactional basis of voicemail, indicating that the callers anticipate continuing their current line of communication with the callee. In the majority of cases, in fact, they anticipate continuing it by phone, as evidenced by their requests for return calls. This sense that voice messages are grounded in a specifi c communicative history helps account for the presence of conversational features in what is at fi rst glance a one-sided speech event. Alvarez-Caccamo and Knoblauch (1992: 481) found that "cultural differences play a very small role in the pattern of messages," (and at least one study comparing the structure of voice messages left in Greek and English found that "a generic pattern for AMMs [Answering Machine Messages] can be established across the two languages" (Goutsos 2001: 357). It may be that the constraints of the medium serve to some extent to unify message-leavers across cultures and languages.

Pedagogical Applications
Nonetheless, personal experience has shown that ESL learners are likely to have diffi culty in planning and executing appropriate voice messages. Non-native speakers may not be aware of the strategies they use and the functions they perform when leaving messages in their native language because they do it automatically. Attention to both the form and function of voice messages can help ESL learners in initially planning messages; and, with enough practice, the skill can hopefully become unconscious. The repetition of phrases like gimme a call, I just wanted to call you (in order) to, and talk to you soon indicates that native speakers have a stock of items that are considered lexically appropriate for voice messages, something ESL learners probably need explicit instruction in. Given the importance today of being able to leave effective and appropriate voice messages, this is a skill that is worth spending time on in the classroom.

Conclusion
The messages gathered show a high degree of uniformity in their macro-structure (OPENING-BODY-CLOSING), as well as in the internal structure of their openings and closings. The content of the bodies of the messages reduces to a small set of discursive functions, and the distribution of several of these functions -justifi cations, call requests, wellwishing -is fairly predictable. The use of specifi c words and phrases to execute certain functions is common across callers, and there is a small set of discourse markers that are used to bound functional units.
In addition, voice messages embody and add to the communicative history of the caller and callee. They display many features that show their interactive foundation, including ritualized greetings and farewells, prosodic marking of the boundaries of functional units, and references to other phone calls both real and anticipated.
Humor seems to appear frequently in voicemail, though it is unclear whether it is a technique for overcoming the irony of talking to a machine, a response to the message prompt, or a simple conversational feature that is not uniquely triggered in voicemail.
These fi ndings could be useful in the design of activities for ESL students learning to leave voice messages. Activities that bring attention to the functions executed and the lexical and grammatical forms therein would be particularly useful.
This study looked at a relatively small sample of messages from a restricted demographic. How voice messages vary in register and micro-content according to such factors as their overall purpose, the relationship between the caller and callee, age, and gender has yet to be thoroughly explored. What is clear, however, is that voice messages, far from being random bits of disorganized speech, are highly structured acts that are grounded in the communicative history of the caller and callee. As voicemail has come to play in increasingly prominent role in more and more people's lives, clear routines have arisen for dealing with the challenges the medium presents. It appears, despite Dear Abby's proclamation three decades ago, that people have learned to speak to machines after all. [32] Hello, this is SC calling from the Association of TJA about the Burgess Scholarship. I'm uh sorry, I didn't get back to you when you phoned last week I was away at a conference. Uh, your application was received, everything is in order. You'll probably get an email confi rmation as well as soon as we um, are, able to get to it. Um, but thank you for your call and please be sure that your application is here and is in good order and is being considered. Thanks so much. Bye. I take it that you remembered the, the daylight, time or whatever, it's not daylight savings time because we lose, lose an hour it's something but, I didn't and I feel like shit. I thought that I was gonna be way early and that I was gonna get there, like to, see the Idiots Karamazov I...like an hour ahead of time, you know that I would get there and I would be able to hang out I'm actually like pretty late at this point. So, I feel like a shithead. I was hoping that somebody else was a shithead too, but apparently everyone else is a good person. So_, anyhow enjoy the show if that's where you are and I hope that your show went well hon. And I'll see you in Blood Wedding.

23
[37] Hey A it's J, umm I'm just wondering if you're going to the Idiots Karamazov tonight [rising tone], I know M talked about umm it, but I didn't know if it was him or you guys...I'd love to meet you guys there, umm. It's in like, a half hour, so I was just wondering if you had M's number if you're not going, but um. I'll probably just head over there and look for you guys. Anyway, bye bye.