This work investigated the language abilities of early language learners from a functional theoretical perspective on language. The dependent variables in the research were the performances of the learners on both a test of abstraction and a probe for acquisition. The independent variable was the use of a multiple exemplar instructional sequence across four different autoclitic frames for spatial relations. The independent variable in the first study extended the multiple exemplar instructional sequence across two response types—listener and speaker. The first study employed a single case, multiple probe design across participants. In the first study, four participants, all diagnosed with language impairments and between the ages of 7-9, demonstrated criterion-level responding (80% correct) to the abstraction test and acquisition probe trials after being taught left, right, on top, and under. The second study employed a single case, multiple probe, delayed multiple baseline design with matched pairs across participants. The independent variable in the second experiment compared the extension of the multiple exemplar instructional sequence across two response types, listener and speaker, with a multiple exemplar instructional sequence across a single response type—speaker. In the second study, four participants, all typically developing, dual language learners, and between the ages of 3-5, demonstrated criterion-level responding (80% correct responding) to the acquisition probe trials after being taught left, right, above, and below. Three participants in the second study demonstrated criterion-level responding (80% correct) to the abstraction test trials. The results demonstrated a functional relationship between the use of a multiple exemplar instructional sequence across frames and the ability of the participants to use autoclitic frames with untaught stimuli. No significant difference in results was found between the use of multiple exemplar instruction across two response types (listener and speaker) when compared to the use of multiple exemplar instruction across a single response type (speaker). The practical and theoretical implications of these findings were both discussed.