Homogeneity as presuppositional exhaustification

Plural definite descriptions give rise to homogeneity effects: the positive The trucks are blue and the negative The trucks aren't blue are both neither true nor false when some of the trucks are blue and some are not, that is, when the group of trucks is not homogeneous with respect to the property of being blue (Löbner, 1987, 2000; Schwarzschild, 1994; Križ, 2015b). The only existing acquisition studies related to the phenomenon have examined children's comprehension only of the affirmative versions of such sentences, and moreover have yielded conflicting data; while one study reports that preschoolers interpret definite plurals maximally (Munn et al., 2006, see also Royle et al., 2018), two other studies report that preschoolers allow non-maximal interpretations of definite plurals where adults do not (Karmiloff-Smith, 1979; Caponigro et al., 2012). Moreover, there is no agreed upon developmental trajectory to adult homogeneity. In this paper, we turn to acquisition data to investigate the predictions of a recent analysis of homogeneity that treats homogeneous meanings as the result of a scalar implicature (Magri, 2014). We conducted two experiments targeting 4- and 5-year-old French-speaking children's interpretations of plural definite descriptions in positive and negative sentences, and tested the same children on standard cases of scalar implicature. The experiments revealed three distinct subgroups of children: those who interpreted the plural definite descriptions existentially and failed to compute implicatures; those who both accessed homogeneous interpretations and computed implicatures; and finally, a smaller subgroup of children who appeared to access homogeneous interpretations without computing implicatures. We discuss the implications of our findings, which appear to speak against the implicature theory as the adult-like means of generating homogeneous meanings.

Parser testing represents the basis of compiler testing because the accuracy of parsing execution directly affects the accuracy of semantic analysis, optimization and object code generation. It should include tests for correct and expected, but also for unexpected and invalid cases. In this paper, techniques for testing the parser, as well as algorithms and tools for test sentence generation are discussed. Generation of negative test sentences by modifying the original language grammar is described. Positive and negative test cases generated by Grow algorithm and Purdom algorithm are applied to the testing of L-IRL robot programming language and obtained results are described in this paper.