The linear ordering Rlex<ω is the lexicographic linearization of the tree of R-valued functions defined on a finite initial segment of ω and ordered by extension. We identify suitable notions of smallness and largeness for linear orderings that embed into Rlex<ω by using tree representations of chains. Specifically, small linear orderings are representable by inversely well-founded trees, and large linear orderings are representable by fully uncountably branching trees. We prove the rather surprising result that all linear orderings embeddable into Rlex<ω are either small or large. This fact sheds some light on the complicated structure of the linear ordering Rlex<ω, and can be useful in applications to utility theory and preference modeling.