All of this work on
<ul>
	<li> <a href="roles.html"> understanding roles in requirements
		engineering</a>,
	<li> <a href="inevitable.pain.html"> the inevitable pain of software
		development</a>, and
	<li> <a href="users.man.spec.html"> user's manuals as requirements
		specifications</a>
</ul>
falls out of empirical observations of software and requirements
engineering efforts over the years.

<p>
In addition, Walter Tichy and I were able to spot threats in work
by A.E.K. Sobel and M.R. Clarkson titled ``Formal Methods Application: An
Empirical Tale of Software Development'', published in <cite>IEEE
Transactions on Software Engineering</cite> SE-28:3, 308--320, March 2002.
They analyzed data to draw conclusions that use of formal methods caused
students to produce more correct software than that produced by students
who did not use formal methods. However it turned out that the students
using formal methods had volutarily learned formal methods. Thus it was not
possible to exclude an alternative explanation for the results, namely that
the better students naturally took the voluntary formal methods class and
these better students naturally wrote more correct programs than the poorer
students. This observation was reported in a note:
<br>
Berry, D.M. and Tichy, W.F., ``Formal methods application: an empirical
tale of software development'', <cite>IEEE Transactions on Software
Engineering</cite>, SE-29:6, pp. 567--571, June 2003
<a href="FTP_SITE/reprints.journals.conferences/critique.new.pdf">
PDF preprint</a>