The SGML and XML standards use a variation of regular expressions
called *content models* for modeling the markup structures
of document elements.
SGML content models may include so called *and* groups, which are excluded from XML.
An *and* group, which is
a sequence of subexpressions separated by an &-operator,
denotes the sequential catenation of its subexpressions in any possible
order. If one wants to shift from SGML to XML in document production,
one has to translate SGML content models to corresponding XML content models.

The allowed content models in both SGML and XML are restricted by a
requirement of *determinism*, which means that a parser recognizing
document element contents has to be able to decide without lookahead,
which content model
token to match with the current input token, while processing the
document from left to right.
It is known that not all
SGML content models can be expressed as an equivalent XML content
model.
It is also known that transforming an SGML content model into an
equivalent XML content model
may cause an exponential growth in the length of the content model.
We discuss methods of eliminating *and* groups and analyze the
circumstances where they can be applied.
We derive a tight
bound of *e n*! on the number of symbols in the result of eliminating
an *and* group of *n* symbols,
where *e* = 2.71828... is the base of natural logarithms.
We present the analysis in a pedagogical manner, emphasizing
mathematical methods which are typical to the analysis of algorithms.
We also show that minimal deterministic automata for recognizing an
*and* group of *n* distinct element names contain *2 ^{n}* states and