The International Phonetic Alphabet (IPA) draws a fundamental distinction between consonants that are voiceless and those that are voiced. For glottal consonants, however, this distinction is problematic. The voiceless glottal stop [ʔ] and fricative [h] are usually realized as creaky and breathy voice, respectively, when they occur between vowels and in weak prosodic positions (Pierrehumbert & Talkin 1992, Ladefoged & Maddieson 1996). Although voicing of [ʔ] and [h] is pervasive, the IPA still distinguishes these sounds from their voiced counterparts, e.g. voiceless [h] vs. voiced [ɦ]. Do voiceless and voiced glottal sounds really differ in terms of their voicing, and if so, how? I will present a phonetic analysis of glottal sounds in “Illustrations of the IPA”– short papers published in Journal of the IPA that illustrate the main sounds of a language with accompanying audio. An analysis of the audio recordings confirms that voicing variation is widespread for both voiceless and voiced glottal sounds. Voiceless [h] is only slightly less voiced than voiced [ɦ], and only in utterance-initial position; elsewhere, both [h] and [ɦ] are as strongly voiced as breathy vowels. Creaky vowels are more strongly voiced than glottal stops, unless the creaky vowels are described as being “rearticulated” or “checked.” Based on these results, I argue that voicing during glottal sounds is largely predictable from respiratory and prosodic factors. In many languages then, glottal sounds can be considered phonetically unspecified for voicing.