An Analysis of Author-Specific Email Signature Styles:
A Case Study

Jeremy Elson
October 21, 1998

I was in a silly mood and had a couple of extra hours on my hands...

From: Jeremy Elson 
To: Mohit Talwar 
cc: dgroup@catarina.usc.edu
Subject: Re: wondering 
In-Reply-To: Message from Mohit Talwar  
   of "Wed, 21 Oct 1998 13:14:45 PDT."  
Date: Wed, 21 Oct 1998 18:01:53 -0700


Mohit Talwar writes:
>why people (a lot of them) sign their emails as
>
>'- name'
>
>I mean, why the '-' ?

Mohit:

Thank you for your interesting and thought-provoking question.  It has
spurred me on to doing extensive research on this topic from my email
archives, the results of which I present below.  Although my research
does not directly answer your question ("why the '-'?"), I think it
does shed some interesting light on that question, as well as
quantifying to some degree your guess of "(a lot of them)".  I would
like to discuss these results in Dgroup very soon so that I can make
the submission deadline for ACM SIGSIG '99 (the ACM Special Interest
Group on Signatures).

(by the way, in case it is not clear -- this is only meant to make you
laugh, it is NOT SERIOUS!!! :) )

------------------------------------------------------------------------


	   Analysis of Author-Specific Email Signature Styles:
				A Case Study

             J. Elson, University of Southern California


ABSTRACT

Email lives in an interesting space halfway in-between informal
conversation and formal written communication.  There are many
stylistic components of an email that are typically not characteristic
of traditional written correspondences: for example, capitalization,
spell-checking, position of quoted material, and other factors.  One
of the most interesting and illuminating is the way in which email
authors sign their names.  A number of different styles are presented,
and their significance discussed.  We conclude that while styles
differ radically across authors, an individual typically uses a
consistent style.


INTRODUCTION

Email has become the primary means of communication in many academic
and technical environments.  As users' familiarity with email
increases, elements of their personalities become incorporated into
the way in which their email is written.  These styles vary in a large
number of dimensions: the selection of a subject, the style of the
greeting, the capitalization conventions used, the rigor of spell
checking, inter-sentence and inter-paragraph spacing, the form of
the signature, and a wide variety of other features that are not
necessarily consistent from author to author.  Our research initially
focused only on the signature; other aspects of email were beyond the
scope of our current interest but are an area of active research.

Our analysis of email signature styles has indicated that email
signatures do not follow an "ad hoc" pattern as has previously been
suggested.  Rather, it is possible to classify them in a way that
exposes an underlying structure.  It is also interesting to note that
signature style variability is quite low when analyzing the set of
emails written by a single author.

Our classification system is presented in the next section.  Section 3
presents our analysis of the most popular forms of signatures.
Section 4 describes our conclusions.


A TAXONOMY OF EMAIL SIGNATURE STYLES

A signature can initially be broken down into two distinct
components.  Consider the following example:

regards,
jeremy

This paper will define the "salutation" to be the generic word that
signifies the imminent end of the email.  In this example, "regards"
is the salutation.  The second component we will define is the
"personal identifier" (abbreviated PI), which is a series of
characters that identify the author of the email.  In the example
above, "jeremy" is the PI.

A wide variety of signature styles were observed in traffic from the
"dphds" mailing list in the time period of this study (August
26-October 21, 1998).  The signatures in our study varied both in
their salutations and PIs.  Our first analysis revealed that there
were an unacceptably large number of distinct styles when both the
salutation and PI were considered; allowing both to vary results in a
combinatoric explosion of possible email signatures.  Therefore, only
the PI was considered; the salutation will be ignored for the
remainder of this paper.

Although the number of PI styles at first seemed to be large, patterns
soon began to emerge which lead to a classification system.  Our
classification system is shown in Figure 1, along with a number of
sample email signatures taken from the sample group to illustrate some
of the classes.



				|Capitalized
				|
				|	|First Initial Only (not full name)
				|	|
				|	|	|Pre-Name Characters
				|	|	|
				|	|	|	|Post-Name Characters
Reference	Example		|	|	|	|
-----------------------------------------------------------------------------
[DEB98]		D.		x	x		x
[AHM98]		-A		x	x	x
[NIR98]		Nirupama.	x			x
[PAV98]		Pavlin		x
[GRA98]		--Graham	x		x
[ANO98]		-Anoop.		x		x	x
[JER98a]	j.			x		x
[JER98b]	j			x
[YIN98]		-ying				x
[MOH98]		mohit.					x
[AMI98]		amit

Figure 1: Email Signature Taxonomy with Examples



Unfortunately, this classification system is not sufficiently
descriptive to capture all possible PI styles.  For example, the PI
"- Haobo" of [HAO98], the "--Graham" of [GRA98], and the "-Polly"
of [POL98], while clearly different, are considered equivalent
because our classification does not distinguish between types and
numbers of pre-name characters.  Also, the fascinating example of
"ReZa" [REZ98], a distinctive style contained only in the writings of
R. Rejaishushtari, does not fall neatly into our system in a way that
distinguishes it from others that are not like it.  Although this
clearly does a disservice to the flair exhibited by the author, a more
complex classification system that did capture such aspects of a style
led to poorer statistical analyses.

Another variation not easily described by our system is the style of
simply not using a Personal Identifier at all.  This is style is
commonly used in the most informal emails; or, when the author relies
on a ".signature" file to provide the distinctive flair that a
signature often provides.  (It should be noted that adding "a personal
touch" is often a signature's only function, as the identification of
the author is typically performed by the email header).

The email of D. Estrin, for example, is most frequently ended without
a PI; the example given by [DEB98], while included in Figure 1 to
provide a more complete view of our taxonomy, is not typical for that
author.   This might lead us naturally to the question: do authors
remain consistent in their use of email signatures?  This question is
an area of future research.


TAXONOMY-BASED STATISTICAL ANALYSIS

198 messages representing 24 different authors were collected from the
"dphds" mailing list during the period of this study.  The raw data
are listed in Appendix A.  Analysis of the frequency of different
signature types is somewhat complicated by the appearance of a large
number of null signatures (71, or 35.8%).  A null signature can either
be considered a valid signature consisting of no characters, or can
alternately be considered a non-signature and inappropriate for
inclusion in these statistics.  Therefore, frequencies will be
reported using both interpretations.

The first question that drew our attention (and, indeed, was the
instigator of the study) was the frequency of use of pre-name
characters.  In our sample, 69 signatures appeared with pre-name
characters -- 54.3% of non-null signatures, and 34.8% of all
signatures.

Authors seem to remain remarkably consistent with their choice of the
use of pre-name characters.  Of the authors who used non-null
signatures, 9 (42.8%) used "-" exclusively; 10 (47.6%) used no "-"
exclusively, and 2 authors (9.5%) are recorded as having used both "-"
and no "-" in separate non-null-signed emails.

It is interesting to note that, while the analysis based on the number
of signatures collected revealed a larger number of signatures
containing a leading "-", an author-centric analysis showed a smaller
number of authors who consistently prefer this style.  This suggests
that "-"-using authors are more prolific writers.

The use of "-" does not appear to correlate with any of the other
three classification types that we previously defined.  For example,
of the 69 signatures that used a "-", 10 (14.5%) did not use an
initial capital letter; similarly, of the 58 non-"-" (and non-null)
signatures, 9 (15.5%) did not use an initial capital letter.

An analysis of the correlation between the use of an initial "-" and
the other two classified types of signatures could not be performed
before the expiration of this project's DARPA funding.  This is an
area of future research.


CONCLUSIONS

The use of an initial "-" before the personal identifier part of an
individual's signature is a popular style.  Although representing only
42.8% of authors, emails signed with a leading "-" account for 54.3%
of non-null signatures.  This suggests that, of authors who choose to
sign their names, those who sign with "-" write more email than those
who do not.  The use of "-" appears not to correlate with other
aspects of a signature's style.

It is also concluded that the time spent to conduct the research could
have been better spent by the author taking a walk on a beautiful
day.  This is an area of future research.


FUTURE WORK

There are a number of interesting research questions whose answers
might serve to further illuminate the culture of email, but were
beyond the scope of our initial research.  For example, it would be
interesting to note the variability of signature use for a single
author.  Most authors do have a preferred signature that is most
commonly used, but the exact frequency of deviation from that
signature for the average author has not yet been determined.

Such a study might lead to the use of email signatures as a potential
(though admittedly weak) authentication mechanism.  Email that is not
signed in a manner consistent with the author might be suspected of
being a forgery.

A number of other future research directions have been proposed,
including an analysis of the "thankful imperative" style, e.g. a
request for some action signed by "Thanks!!", suggesting that the
request is urgent and requires immediate attention. (Initially
suggested by Dr. L. Girod.)


ACKNOWLEDGMENTS

The email archives used in this study were available in part due to
computational resources made possible by an NSF Small-Scale
Infrastructure Grant, award number CDA-9216321.  Dr. M. Talwar
originally suggested these research directions.  Additional thanks go
to Dr. L. Girod and Dr. P.I. Radoslavov for useful discussions and
reviews of early drafts of this manuscript, and Dr. D. Estrin for
the creation of the mailing list which formed the centerpiece of this
study.


REFERENCES

[AHM98] Ahmed A-G Helmy, "micronets", 14 Oct 1998, 11:40:10 PDT
[AMI98] Amit Kumar, "About connection speeds of a modem", 22 Sep
	 1998, 19:32:51 PDT
[ANO98] Anoop Reddy, "Re: who is on this week?", 12 Oct 1998, 9:54:20
	 PDT
[DEB98] Deborah Estrin, "Re: Dgroup wed 9th (fwd)", 8 Sep 1998,
	 10:50:41 PDT
[GRA98] Graham Phillips, "Re: egcs", 20 Oct 1998, 11:20:12 PDT
[HAO98] Haobo Yu, "mbone tomorrow", 20 Oct 1998, 15:01:03 PDT
[JER98a] Jeremy Elson, "Security slides available on the web", 16 Oct
	 1998, 16:18:57 PDT
[JER98b] Jeremy Elson, "homework 1 turned in today", 20 Oct 1998,
	 16:23:21 PDT
[MOH98] Mohit Talwar, "vern paxon's lecture", 3 Oct 1998, 14:12:18 PDT
[NIR98] Nirupama Bulusu, "wednesday talk", 19 Oct 1998, 17:36:52 PDT
[PAV98] Pavlin Ivanov Radoslavov, "Re: egcs", 21 Oct 1998, 11:10:54 PDT
[POL98] Polly Huang, "Re: Memory exceeded", 19 Oct 1998, 10:21:11 PDT
[REZ98] Reza Rejaishushtari, "Re: Memory exceeded", 19 Oct 1998,
	 10:16:39 PDT
[YIN98] Ying Liu, "Re: dgroup", 21 Oct 1998, 10:00:39 PDT


APPENDIX A: Signature Frequencies

     18	- Haobo
      1	- Lew
      2	- haobo
     19	--Graham
     10	-A
      1	-Ahmed
      2	-Anoop.
      4	-Polly
      1	-Puneet.
      1	-Ya
      1	-Ying
      1	-Ying Liu
      1	-amit
      6	-ying
      1	-ying liu
      1	Cengiz
      4	Chalermek
      2	D.
      1	Eddie
      2	Niru
      1	Niru.
      1	Nirupama.
     27	Pavlin
      6	ReZa
      2	Satish
      2	Ya
      1	amit
      2	art mena
      6	mohit.
     50	 (D. Estrin)
     21	 (Other authors)

Back to my writings page
Back to my home page