In this dissertation, I examine how knowledge is communicated through speech in Kyrgyz based on 15 months of ethnographic fieldwork from May 2018 to September 2019 with horse herder and horse trainer families in Teskey Kojoke, Kyrgyzstan. Specifically, I examine Kyrgyz herders’ and trainers’ everyday interactions with their horses, and the chains of semiotic encounters through which they make knowledge claims and give accounts of their own and others’ claims.One aspect of how knowledge is communicated involves the source of knowledge claims. According to traditional accounts in linguistics, Kyrgyz, like other Turkic languages, exhibits evidentiality, meaning that the source of the speaker’s knowledge is communicated through grammatical marking via morphemes. In such accounts, morphemes mark sensory inputs that are linked to mental state constructs. In this dissertation, I demonstrate that the common linguistic model of grammatical evidentiality does not reflect how Kyrgyz speakers convey information about the source of knowledge. Specifically, in Kyrgyz, configurative templates make categorial distinctions that are confused with the function of single morphemes. Furthermore, the analysis of categorial distinctions indexed by configurative templates reveal that evidential construals cannot be specified by sentences. Therefore, I demonstrate that sensory inputs are not categorial distinctions marked by morphemes in sentences. I showcase that evidential construals - the source of speakers’ knowledge claims - can only be recovered if we attend to the deictic anchoring between co-textual arrays and categorial effects. In fact, this recovery requires ethnographically documenting relationships between propositional acts across chains of semiotic encounters. This reconstitution of evidentiality reveals a distinct organization of experience for Kyrgyz speakers based on who is the witness to which semiotic events. Since such alignments across events shift as people interact with each other on different occasions, ‘knowledge’ has ever-shifting qualities based on who the speakers and addressees are and the relationship between the current speech event and the narrated event. By ethnographically documenting relationships between propositional acts across chains of semiotic encounters, I reconstruct ‘knowledge’ as a semiotic process that is worked-out by individual practitioners in routinized inferential processes.