Theories of grammars, or the mapping between sounds and meanings, are typically constructed on the basis of patterns of acceptable sound-meaning pairings. However, theories of grammars also define the space of possible representations the parser can construct in real time, and make commitments about what information the learner must be able to detect in her primary linguistic data. Rarely are these consequences of specific grammatical accounts pursued carefully. In this talk, I take two controversial topics — the that-trace effect and the status of resumptive pronouns in English — and show how data from the learner’s input and real-time processing delimits the space of possible accounts of these phenomena, given plausible assumptions about UG and grammar-parser relationships. I show that accounts of the apparent inapplicability of the that-trace effect in Romance languages imply that learners must track the distribution of complementizers and extraction sites (e.g., Rizzi & Shlonsky '06), whereas other theories increase the range of possible sentences that count as "successes" (e.g., Rizzi '82, Chomsky '15). I show that the distribution of complementizers and extraction sites in English, Spanish, and Italian child-directed speech are likely not distinct enough to account for the differences in the adult grammars. Thus, the contrast in the applicability of the that-trace constraint must be learned "indirectly", i.e., it must be conditioned on some other detectable grammatical property, such as subject positioning or richness of agreement. Additionally, I show that processing data and learnability concerns inform our accounts of resumptive pronouns in English and Hebrew. Some accounts hold that resumptive pronouns in English are grammatical repair strategies inside island contexts (e.g., Ross '67), whereas later accounts claim that English resumptive pronouns are "intrusive", or extragrammatical repair mechanisms (e.g., Sells '84). This is typically in contrast with languages like Hebrew, which are typically described as having 'grammaticizied' resumption. However, carefully collected judgment data between the two languages paints an unclear picture – resumptive pronouns are typically rated lower than gaps in English (Keller & Alexopoulou '07, Heestand et al '11), but are also rated very lowly both inside and outside of islands in Hebrew (Farby '10). We show that Hebrew- and English-speaking adults show different processing profiles for resumptives in islands. In real-time sentence processing, Hebrew-speakers actively search for resumptive pronouns inside island contexts, whereas English-speakers do not. If the parser only actively searches for parses that are sanctioned by the grammar, then these results follow if the English-Hebrew contrast inside island contexts is ultimately a grammatical distinction. Additionally, we show that the distribution of resumptive pronouns in English and Hebrew child-directed speech are only distinguishable outside of islands, and thus accounts of the acquisition of resumption in languages like Hebrew must treat resumptive pronouns inside and outside of islands as a natural class, licensing inference from distribution of resumptive pronouns outside of islands to the acceptability of resumptive pronouns inside islands. This has consequences for accounts of resumptive pronouns which do not treat resumption inside and outside of islands as a natural class.