This dissertation makes two contributions, one empirical, the other theoretical. Empirically, the dissertation deepens our understanding of the lifecycle and behavior of the preaspirated stop, an extremely rare phonological feature. I show that in most confirmed cases, preaspirated stops develop from earlier voiceless geminate stops, less commonly from nasal-voiceless stop clusters. When decaying, preaspirated stops typically develop into unaspirated voiceless stops, or undergo buccalization to become preaffricated. More rarely, decaying preaspirated stops may trigger tonogenesis, or undergo spirantization or nasalization. Phonologically, preaspirated stops usually function as positionally conditioned allophones of underlying aspirated voiceless stops contrasting with voiceless unaspirated stops. The dissertation tests three theoretical frameworks. First, the State-Process model (Greenberg 1978, 1969, 1966) claims that the synchronic distribution of linguistic features offers insight into their rates of innovation and transmission. Conventionally, the rarity of preaspirated stops is attributed to a presumed low rate of transmission: they are rare because they are hard to hear (Silverman 2003, 1997; Bladon 1986). However, the geographic and genetic distribution of preaspirated stops fit the State-Process model's prototype of an infrequently innovated but robustly transmitted linguistic feature. Further, I show experimentally that preaspirated stops are no more difficult to distinguish from unaspirated stops than are much more abundant postaspirated stops. Second, the dissertation tests the success of two models, one cognitive, the other phonetic/diachronic, at accounting for two place-based asymmetries in Scottish Gaelic preaspiration. Whereas a conventional Optimality-Theoretic analysis of these asymmetries overgenerates, an analysis modified via Steriade's P-map Hypothesis (2001a, 2001b) resolves this overgeneration. The P-map analysis depends on congruent perceptual scales, which the perception experiment (above) confirms: participants' confusion rates closely match the place-based asymmetries observed in Gaelic. The competing "innocent misperception" model (Ohala 2005, 1993; Blevins 2004) depends on the presence of phonetic precursors to produce an ambiguous phonological signal, which listeners may interpret differently than intended by the speaker, leading to an alteration in a segment's underlying form. A series of production experiments identifies potential precursors, but also reveals between-speaker variation more compatible with the P-map account than "innocent misperception," again lending support to Steriade's hypothesis. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]