This dissertation examines three processes affecting consonants in child speech: harmony (long-distance assimilation) involving major place features as in "coat" [kouk]; long-distance metathesis as in "cup" [p[wedge]k]; and initial consonant deletion as in "fish" [is]. These processes are unattested in adult phonology, leading to proposals for child-specific constraints. Initial consonant deletion in particular is a little-understood phenomenon thought to be idiosyncratic. However, my survey of initial consonant deletion as reported in eight languages reveals systematic deletion patterns affecting continuants and sequences of different consonants. I argue that all of these child-specific processes are tied to the acquisition of consonant sequences. In order to understand the role of these processes in phonological development, I examine consonant acquisition data from a diary study of Grace, an English-acquiring child. I adopt the Bernhardt and Stemberger (1998) variant of Optimality Theory for the analysis since their view of default underspecification, sequences of features, and feature-based approach to sonority permit a unified analysis of harmony, metathesis and initial consonant deletion that explains Grace's trajectory of acquisition as well as the frequency of certain patterns across children. I show that independently motivated constraints governing feature sequences, onset sonority preferences, initial velars, and the tendency to anticipate features within a prosodic domain explain all of these processes, as well as Grace's onset cluster reduction patterns (e.g. "snake" [seik]) and gradual acquisition of different cluster types. Children must learn to produce consonant feature sequences within a word before producing sequences within an onset. Child-specific processes are eliminated as children acquire the speech planning skills necessary to express the contrasts of a mature language, though the constraints remain active in adult phonology. The longitudinal data provide evidence for both constraint demotion and promotion in learning, as well as distinct roles for two types of faithfulness constraints. One mandates the preservation of non-default features that are specified in the underlying representation, while the other evaluates identity of a correspondent segment to any non-default feature associated with a segment. This distinction permits the derivation of initial consonant deletion as a response to positional constraints on features or feature sequences. [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.]