The notion of a 'stem group' is indespensible for a palaeontologist. Much used and abused, it is simply not possible to talk about the relationships of fossils to modern life without the use of the crown and stem group concepts. The crown group is a clade which is delimited by its living (extant) members. The stem group comprises those fossils which are closer to the crown group than to any other extant clade, but do not fall within the crown group. As a result, the stem group is paraphyletic, and thus not really a group at all. It is perhaps more useful to talk about a 'stem assemblage' than a 'stem group'.
While at this year's SVP (and at previous meetings), I was struck by some of the terminological abuses of the term 'stem group'. In various instances, it was used either to refer to the nearest sister taxa of an extinct clade, or it appealed to essentialist nomenclature. I comment further on these below the fold.
'Stem groups' of extinct clades:
When a clade is extinct is has neither a crown nor a stem. If we did not distinguish between extant and extinct clades when applying the crown group concept, then crown groups could be arbitrarily small and stem groups arbitrarily deep. Because nodes in a cladogram are rotatable, we could use any taxon (fossil or living) to be a stem taxon.
We already have a set of terms for this: sister group relationships. This is also what the crown group concept conveys. However, it's purpose is to convey the relationship of fossils to a particular living group. When we talk about fossil or extant clades, we can talk about the nearest sister taxa. When talking about fossils in relation to an extant clade, only then do we apply the crown group concept.
Arbitrarily deep stem groups
One abstract title at this year's meeting struck me, because it referred to the fossil Morganucodon as the earliest stem-mammal. This taxon is almost certainly a stem-mammal. Is it the earliest? Take a look at this figure (from Angielczyk, 2009) (you may have to click on it to see the full image):
Notice the placement of the node "Mammalia". It's a full two internodes displaced from the node that subtends the extant mammalian branches: monotremes, marsupials, and placentals. You'll also notice that the Triassic fossil Morganucodon is the nearest fossil sister group of the three extant mammal lineages. In other words, it's the nearest sister taxon (in this tree) of the mammalian crown group (which, strangely, is unnamed!).
This is a peculiar trait among palaeontologists: give the standard crown group name (i.e Mammalia, Aves, etc.) to some arbitrary node within the group's stem. For instance, Aves (birds) is often considered to be the clade delimited by the last common ancestor of all extant birds + Archaeopteryx.
What you should also notice in the diagram above is that the root node of this tree is called "Synapsida". This means it entire run of taxa in this tree from the Synapsida node up to (but not including) the unnamed mammalian crown group nodes are part of the mammalian stem assemblage. Yes, Dimetrodon is a stem mammal, as well as Morganucodon. This means that a host of Permian (and potentially earlier) forms are also stem mammals, leaving Morganucodon appearing fairly late in the game.
The utility of the stem/crown group concept comes in placing fossils in relation to living groups. When we do this, fossils can be used to build up knowledge of the sequence of acquisition of homologies where living forms provide no clues. Fossils can, in turn, help test hypotheses of homology by providing unexpected combinations of characters, as well as precluding or 'predicting' certain character combinations. It is important that these concepts are applied in the correct fashion, or else they (and fossils) will lose their meaning.
Angielczyk, K. 2009. Dimetrodon Is Not a Dinosaur: Using Tree Thinking to Understand the Ancient Relatives of Mammals and their Evolution. Evolution: Education & Outreach 2:257–271.