This paper presents a model of second-position clitic (2PC) placement in two Salish languages, St'át'imcets (Lillooet, ISO 369-3:lil) and ʔayʔaǰuθəm (Comox-Sliammon, 369-3:coo). Under the proposed analysis, clitic linearization occurs post-syntactically, but at a point where syntactic structure is still partially accessible, in the form of a span (a complement sequence of heads: see Svenonius 2012). Linearization proceeds bottom-up: each 2PC is lexically associated with an enclitic feature which forces it to attach to the next highest head in the span. If that head is another 2PC, it too will need to cliticize to the next highest head, creating a "clitic string" as a morphological constituent. If there is no higher head, the entire string then encliticizes to the closest available non-clitic head below it in the span, creating the "second position" effect. Crucial evidence for this analysis comes from coordination. In coordinate structures (including those with more than two conjuncts), the clitic string may appear following the initial word of any of the conjuncts, but may not be split across conjuncts. Using the labeling algorithm proposed for coordination by Chomsky (2013), the analysis treats the head of each conjunct as equidistant to the clitic string, allowing for variable placement. In contrast, neither a syntactic nor a phonological account of 2PC placement can account for these coordination facts.