Most current models of sentence comprehension assume that the human parsing mechanism (HPM) algorithmically computes detailed syntactic representations as basis for extracting sentence meaning. These models share the assumption that the representations computed by the HPM accurately reflect the linguistic input. This assumption has been challenged by Ferreira (2003), who showed that comprehenders sometimes misinterpret unambiguous sentences in which subject and object appear in noncanonical order, such as passives or object-clefts. According to Ferreira, these misinterpretations show that parallel to an algorithmic analysis, the HPM performs a heuristic analysis sometimes resulting in interpretations not licensed by the grammar. Our study investigated whether misinterpretation effects indeed reflect an erroneous mapping of form to meaning due to heuristic processing strategies. Using an experimental design closely following Ferreira (2003), Experiment 1 demonstrates that errors with noncanonical sentences show up in German as well, despite the fact that German provides morphological case, which a heuristic strategy should use. Experiment 2 required participants to judge the plausibility of the same sentences. With this task, no evidence for misinterpretation of noncanonical sentences was found. Taken together, our results suggest that misinterpretation errors do not reflect errors in the mapping of form to meaning, but task-specific difficulties that arise when participants retrieve information from the memory representation of a sentence. Consequently, misinterpretation errors do not provide evidence for the claim that the HPM pursues a heuristic analysis in addition to an algorithmic analysis. Our results instead lend support to models of the HPM that assume algorithmic processing only.