Non-target patterns of language production in heritage and L2 populations stand to reveal a great deal about the unity of behaviors under ideal resource availability (competence) and capacity-limited behaviors (performance) of these grammars. In this presentation I present on going collaborative research (Hopp, Putnam, & Vosburg; under review) where we investigate whether non-target wh-questions in heritage Mennonite Low German (a.k.a. Plautdietsch) and L2 English speakers are due primarily to cross-linguistic transfer or the reduction of grammatical complexity as modeled by the Derivational Complexity Hypothesis (DCH, Jakubowicz, 2005). Previous research shows that complex (i.e., cross-clausal) wh-dependencies pose more difficulty to child L1 and adult L2 learners than monoclausal dependencies (Jakubowicz & Strik, 2008; Schulz, 2011; Slavkov, 2014). In this study, we investigate the linguistic behavior of twelve (n=12) bilingual Plautdietsch-English speakers in Southwestern Kansas, analyzing both their production and comprehension of wh-questions in both languages. In both production and comprehension, in the L1 only heritage speakers produced medial, non-target wh-elements, while in L2 English, only late L2 learners produced such elements. We argue that these patterns cannot be due to transfer, since speakers produce medial-wh in only one of their languages. Instead, medial wh-elements occur as a mechanism to reduce syntactic complexity in the less dominant language, irrespective of whether it is the L1 or L2 and regardless of whether the language was acquired early or later in life. These findings suggest that the DCH can account for aspects of grammatical restructuring in both L1 and (late) L2 speakers.