Three studies demonstrate that aspects of English blend structures lik
e ''brunch'' are quite predictable from cognitive and linguistic princ
iples The first study analyzed the order in which words are represente
d in blends and found that thr ee variables that predict word order in
conjunctive phrases like ''smoke and fog'' also predict the order in
which word components appear in blends like ''smog.'' In particular, t
he first word represented in blends tends to be higher in frequency, c
ontain fewer syllables, and denote more prototypical category members
than the second word. The second study examined the boundaries between
blend components and found that they fall primarily at major phonolog
ical joints, such as syllable, onset, and rime boundaries. Breaks at o
nset-rime boundaries were more than four times more common than breaks
at body-coda boundaries, thus supporting an onset-rime representation
of English syllable structure. The third study showed that the phonem
es involved at the juncture between blend components tend to be phonol
ogically similar, which might be used to emphasize the teasing nature
of blends. More generally, the studies illustrate the promise of takin
g hypothesis-testing approaches to blend structure rather than the tax
onomic approaches common in prior investigations.