A) Experimental studies of language comprehension processes in adults

The central question is what the mental representations of linguistic knowledge are, how this knowledge is used in language comprehension, and what kinds of mental representations are built during the comprehension process.

From a methodological standpoint, it is essential to carry out experimental investigations that allow the comprehension process to be studied as it unfolds, in real time, as linguistic information is perceived. To this end, a laboratory for the real-time analysis of reading processes was set up, using experiment-control software able to present a text visually or auditorily, one word or one sentence at a time, and to record subjects' reading and response times with millisecond accuracy.

A first line of research explored what kinds of representations are built by the language parser, by studying the comprehension of sentences with a non-canonical structure, that is, with displaced elements, such as interrogative sentences, and of sentences with post-verbal subjects. The hypothesis (De Vincenzi, "Syntactic parsing strategies in Italian", 1991, Kluwer Academic Publishers) is that during real-time language processing the human parser, given the limits of short-term memory, uses parsing strategies based on principles of economy of derivation and representation: this hypothesis is formulated as the Minimal Chain Principle.

De Vincenzi proposes that the language parser tries to complete these dependency relations as quickly as possible and to form dependencies that are as simple as possible (that is, shorter and with the fewest possible components). This hypothesis has been supported by a large body of experimental data on the interpretation of different types of sentences, both interrogative and declarative.

Another line of research explored parsing strategies in the interpretation of elements within a single sentence constituent. This is the case of modifier attachment, such as the relative clause che si è ferita ('who got hurt') in sentence (2), which can be interpreted as referring either to the noun signora ('lady') or to the noun figlia ('daughter').

The Late Closure strategy, also based on the limits of short-term memory, predicts that the initially preferred attachment should be to the linearly closest constituent.

The experiments conducted by De Vincenzi measured two dependent variables: the first was the reading time of each single word, that is, a measure of real-time comprehension processes; the second was an off-line measure, namely the time needed for the final comprehension of the whole sentence.

The results showed very clearly that on-line, real-time measures reflect the predominance of syntactic processes, which are automatic and unconscious, whereas off-line measures show the predominance of pragmatic factors, which are neither automatic nor unconscious and require access to a very large and variable amount of information.

The importance of this result is twofold: on the one hand, it confirms the validity of the Late Closure strategy in the initial interpretation of Italian sentences; on the other hand, it provides important information about the organization of the human mind, showing that sources of information with a limited and fixed vocabulary, such as syntactic information, are organized in a modular way, that is, they are used automatically and unconsciously and take priority over other types of information.

Another aspect investigated experimentally is how the antecedent of a pronoun is identified, in particular how morphological and syntactic information is used in activating antecedents. The results showed a representational and temporal difference between the use of morphological gender information and that of number information.

Finally, a further aspect is the comparison between the performance of subjects without language disorders and agrammatic subjects. This comparison concerns the comprehension of interrogative sentences, a central topic of De Vincenzi's research with unimpaired adults. This research is important both for a more accurate definition of the agrammatic syndrome (Broca's aphasia), that is, to establish which aspects of the linguistic deficit are due to a selective loss of grammatical knowledge and which to input-processing problems related to short-term memory, and to verify whether the distinction postulated at the linguistic level is reflected at the level of selective brain damage.

B) Experimental studies on developmental aspects of language comprehension

De Vincenzi's research in this field seeks to establish a meeting point between experimental psycholinguistic work with adults and work on language comprehension in developing children. Moreover, an accurate definition of what constitutes a complexity factor in language comprehension is essential to diagnose linguistic disadvantage precisely, and therefore to be in the best position to design targeted remedial interventions.

A first line of research explored the hypothesis that what is merely a complexity factor for the adult reader/listener without language disorders may constitute a linguistic deficit in subjects with language disorders or in developing children.

To this end, the processing of sentences that are lexically very simple but differ in syntactic structure was studied in five-year-old children. The results show a strong correlation between the performance of children and that of adults: in particular, the cases of syntactic complexity that in adults lead to non-preferred interpretations or longer reading times cause errors in children.

Another important aspect of developmental research is the exploration of the relationship between working-memory capacity and the structural complexity of sentences. Also related to language development is a study on the experimental evaluation of different methods of teaching reading.

C) Cross-linguistic comparison between typologically different languages, to highlight the universals of the mental mechanisms of language

These issues were addressed by adopting a comparative approach to the study of comprehension mechanisms in English and Italian, two languages with important structural differences.

The cross-linguistic aspect is particularly important considering that the vast majority of research on these topics consists of experimental work on English. Hence the empirical need to establish whether the conclusions drawn from this work also apply to languages such as Italian, whose structures differ from those of English, with far fewer restrictions on word order. Experiments on the comprehension of interrogative sentences in Italian and English show that the same Minimal Chain Principle is adopted in both languages, confirming the universalist hypothesis of parsing strategies.

In addition to her collaborations with American scholars, De Vincenzi carried out a series of studies with Spanish and French colleagues in the area of Romance languages. The same linguistic material was translated into the different languages, and questionnaires were drawn up to assess whether the preferred interpretation of the material was the same across languages, for equal structural complexity. The same material was then administered as a computerized presentation, to see whether some experimental differences found between Spanish and Italian were due to linguistic differences or to differences in experimental methodology.

The results show that, with the same experimental methodology, that is, the same segmentation of the input and the same type of linguistic material, the same parsing preferences are found, as well as the same dissociation between initial (on-line) and final (off-line) interpretation.