Recent usage-based models of language acquisition research has found that three frequency manipulations; (1) skewed input (Casenhiser & Goldberg 2005), (2) input consistency (Childers & Tomasello 2001), and (3) order of frequent verbs (Goldberg, Casenhiser, & White 2007) facilitated construction learning in children. The present paper addresses the generalizability of these three frequency effects in adult second language construction learning. The study replicates and extends Casenhiser and Goldberg (2005) with an adult L2 learning population (N = 137) in a 2 ([plus or minus]skewed) x 2 ([plus or minus]consistent) x 2 ([plus or minus]forward skewing) factorial design. It was found that there were no significant positive effects of input skewedness on comprehension of an APPEARANCE construction (Casenhiser & Goldberg 2005) or a Samoan ergative construction (Robinson 2005), and had a negative effect on guided production of the APPEARANCE construction. On the other hand, input consistency had significant negative effects on comprehension of both constructions. In addition, it was only on comprehension of the APPEARANCE construction that providing verbs with relatively high token frequency during training had a positive effect. There were no significant interactions among the three frequency manipulations. (Contains 8 tables, 1 figure and 11 footnotes.)