Early Scope Assignment: Epistemic modals and negation in Italian L1 Vincenzo Moscati (Macquarie University) Sentences with multiple logic operators are in principle ambiguous between different alternative readings and this poses a non trivial challenge for children, who need to converge on the same range of interpretations allowed by adult grammar. Initial findings, based on children’s interpretation of sentences with negation and a nominal quantifier (Musolino 1998), suggested that children lack inverse scope readings. This set of phenomena has been grouped under the label of isomorphism, capturing the idea that early logic scope coincides with surface C-command (Musolino et al. 2000). This account has been successively challenged by new empirical findings (Gualmini 2004) and alternatives to isomorphism are based on semantic (Crain et al. 1994, Moscati and Gualmini 2008) or pragmatic factors (Husley et al. 2008). In this paper we will present the results from a new experiment on children’s interpretation of negative modal sentences in Italian. In this language the modal paradigm is unspecified for polarity but sensitive to surface structure order: an ideal testing ground to evaluate an account based on surface C- command against the predictions of the Semantic Subset Principle (Crain et. al 1994). Consider the sentence in (1), where the only possible interpretation is the surface scope reading in (1a). (1) Ci puo non essere una mucca nella scatola there might not be a cow in-the box SCOPE SEMANTIC STRENGTH a. it is possible that there isn’t a cow in the box SURFACE WEAK b. *it is not possible that there is a cow is in the box INVERSE STRONG Interpretation (1a) is target-consistent and isomorphic: proposals based on input frequency and surface C-command expect interpretation (1a) to be unproblematic for children. However, this reading is semantically weak given that it is asymmetrically entailed by the alternative (1b). For this reason, accounts based on the Semantic Subset Principle make instead exactly the opposite prediction: interpretation (1b) should be favored since it has stronger truth conditions than (1a). We investigated sentences as (1) adopting a reasoning scenario based on the hidden object paradigm combined with a Truth Value Judgment Task (Noveck 2001). 23 Italian monolingual children between 5;0-5;11 (mean 5;3) and an adult control group (N=10) took part to the experiment. Subjects were presented with six sets of boxes and in each set, two boxes (A and B) were open while the third one was closed (box C). Subjects heard a short story and they were told that the content of box C was identical to the content of one of the two closed boxes (A or B). An extra object outside the boxes was added in order to balance the materials. Given the context illustrated in fig.1, sentence (1) is true under the adult/surface scope reading (1a) while it is false under the alternative strong reading (1b). Other 5 additional experimental conditions (see Tab.1.) were added in order to control for the (i) associated answer (True or False), (ii) sentence polarity (positive or negative) and (iii) position of negation (pre-modal or post-modal). Results are reported in Table 2. In general, children commit fewer mistakes in positive sentences (conditions 5-6) than in negative ones (conditions 1-4). This is expected, assuming an extra processing cost associated with negation (Wason 1965, Slobin 1966, Staab 2007). Let us now look at negative sentences (1-4). Condition (1) stands out as being the more problematic: it is correctly judged as true only in the 38% of cases by 5 years olds. Notice that this is the only condition where the truth-value of a (deviant) strong reading is different from the correct reading. This result can be readily explained if we assume that children misinterpret (1) in accordance to the reading (1b), false in the given context. This is in line with the predictions of the Semantic Subset Principle and unexpected under alternative theories based on surface C-command or frequency.