Nominalizações infinitivas e regressivas no português brasileiro: uma abordagem formal em interface com a perspectiva experimental

Abstract

This study investigates two types of zero nominalization in Brazilian Portuguese (BP), infinitival nominalizations (e.g., cantar – o cantar, ‘to sing – the singing’) and back formed nouns (e.g., cantar – canto, ‘to sing – song’), by combining a formal approach, grounded in Distributed Morphology (Halle & Marantz, 1993; Marantz, 1997), with an experimental perspective. Our starting hypothesis is that zero nominalizations are subject to the same syntactic mechanisms that licence concatenative forms, not requiring specific resources such as morphological conversion or a formal regression operation to derive these forms. Empirically, the study provides a detailed characterization of infinitival and back formed nominals in dialogue with classic typologies. Building on Chomsky’s (1970) distinction between gerundive, mixed, and derived nominals, we argue that BP infinitival nominals share properties both with gerundive and with mixed nominals, while back formed nouns pattern with derived nominals. Drawing further on Grimshaw’s (1990) typology and, more specifically, on Borer’s (2014a) reinterpretation distinguishing argument-structure nominals (ASN) and referential nominals (RN), we propose that BP infinitival nominals behave as ASN, whereas back formed nominals are ambiguous between ASN and RN and, in certain cases, instantiate a third, hybrid type not predicted in Borer’s (2014a) framework, combining properties of both ASN and RN. To complement this empirical discussion, a corpus study was carried out using data from Linguateca, allowing us to map the distribution of infinitival nominalizations in naturalistic contexts and to delimit their specific properties. From a formal perspective, following Alexiadou, Iordächioaia & Schäfer (2011), we argue that the differences between infinitival and back formed nominalizations are accounted for by the presence or absence of specific functional layers in their syntactic structures. Infinitival nominalizations that behave as ASN display the verbal layers Asp, Voice, and v, and in the nominal domain, the heads Class and D. Back formed nominals that pattern as ASN contain only v and Voice heads, while those identified as RN show merely the verbal categorizer as part of the extended verbal projection. An intermediate group – exhibiting both ASN and RN properties – is characterized by v lacking selectional features and by a Voice head with no argument in its specifier. In all cases, the nominal spine of back formed nominals is uniform, comprising the functional layers D, Number, and Class. On the experimental side, we investigated the processing of these formations through two complementary studies. First, we conducted a maze task (self-paced reading) manipulating two independent variables: (i) type of zero nominalization (infinitival vs. regressive) and (ii) argument-structure type of the verbal base (unaccusative, unergative, transitive), with reaction times and target choices as dependent variables. Although we predicted higher acceptability and shorter reaction times for infinitivals, the results revealed no statistically significant differences. Second, we ran an acceptability judgment task using a four-point scale (1–2 = higher strangeness; 3–4 = higher acceptability). Overall, infinitival nominals with unaccusative bases received higher acceptability ratings than their back formed counterparts, confirming the hypothesis that back formed nominals display gaps. In the transitive condition, the type of nominalization did not affect acceptability: both infinitival and back formed nouns were predominantly judged as odd (levels 1–2), a result we attribute to the explicit presence of the by-phrase, an element not obligatory in this type of construction. Finally, in the unergative condition, acceptability judgments were more balanced, which we interpret as possible evidence of a different status for the agent introduced by the preposition “de” in unergative nominalizations, as opposed to the agent of transitive nominalizations, which is introduced via a by-phrase.