Accurate diagnosis and characterization of voice disorders requires an understanding of typical voicing: how is variation in voice quality used to convey linguistic information in typical speech? This dissertation addresses this question by examining how speakers produce and perceive creaky voice that is phonologically conditioned. ‘Creaky voice’ is a broad umbrella term used to describe voice qualities that have increased constriction, lower pitch, and/or more irregularity relative to modal voice. In English, creaky voice is typically used to convey segmental and prosodic information, but this same quality may be used to characterize dysphonic speech. Importantly, it is unknown if dysphonic and typical voices differ in their use and realization of creaky voice. Three primary research questions are explored: (1) How do typical speakers vary their voice quality to target phonologically-conditioned creaky voice? (2) How do listeners extract linguistic information from variation in voice quality? (3) How do speakers with voice disorders target phonologically-conditioned creaky voice? In Chapter 1, I present evidence that variation between a speaker’s modal and creaky voice is highly structured and predictable, despite wide cross-speaker variation in the acoustics of both qualities. As a result, each speaker produces phonologically-creaky voice in a unique way while aiming for the same target voice quality. In Chapter 2, I examine how listeners cope with this variability, and present evidence that suggests listeners can normalize to a person’s long-term voice quality when recognizing words. Finally, in Chapter 3, I present a case series of two speakers with voice disorders to examine how creaky voice might differ between typical and disordered voice quality. I conclude that production of phonologically-conditioned creak does not necessarily differ from typical speech based on acoustic measures typically used in a clinical setting, at least for speakers with mild dysphonia. Overall, the framework adopted in this dissertation provides a way to compare typical and disordered voices in novel linguistically-principled ways.