Skip to content
BY 4.0 license Open Access Published online by De Gruyter Mouton August 3, 2022

Italian Verbs with Two Auxiliaries: A Forced-Choice Experiment

Stefano Rastelli
From the journal Probus


In this experiment, 62 university students, native speakers of Italian living in Northern Italy, were asked to choose between two nearly identical sentences featuring intransitive verbs which, according to the literature, can take either auxiliary (AUX) essere “be” or avere “have” in compound tenses. A binary logistic regression showed that for 40% of the time, participants preferred one AUX over the other. In general, the frequency of verbs and association scores between the AUX and the past participle in the input straightforwardly determined the remaining preferences, and especially the choice of essere. The impact of verb semantics on participants’ choices depended on the AUX. In the presence of telic completions and inanimate subjects, the likelihood of AUX essere to be chosen increased, whereas no effect of animacy and completion type was found for the AUX avere. Based on these data, one may conclude that (a) AUX essere is more permeable to both frequency effect and verb semantics than avere; and (b) among educated young adult native speakers living in Northern Italy, there is much less optionality in the choice of the AUX than it is assumed by some linguists.

1 Introduction: Topic and Issues[1]

Written spoken contemporary Italian language features about 100 intransitive verbs that can take either auxiliary (AUX)[2] avere “have” (A) or essere “be” (E) in compound past tenses (such as passato prossimo). The examples below feature the same verb accompanied by AUX avere (1) and essere (2) (AUX bolded):

Giovanni ha inciampato
Giovanni have.aux.3s stumbled
‘Giovanni stumbled’
Giovanni è inciampato
Giovanni be.aux.3s stumbled
‘Giovanni stumbled’

Double auxiliary verbs (DAVs) include correre “run”, piovere “rain”, bruciare “burn”, squillare “ring”, appartenere “belong”, cominciare “begin”, durare “last”, emigrare “emigrate”, finire “stop, end”, volare “fly”, seguire “follow”, saltare “jump”, vivere “live”, nuotare “swim”, emergere “emerge”, cedere “give in”, and intervenire “intervene”.

According to reference grammars of modern Italian, AUX A is used when the grammatical subject of the sentence is the agent, i.e., it controls the action, whereas E is used when the grammatical subject is theme, i.e., it undergoes a change of state or position (Maiden & Robustelli 2000: 263–268). Other authors suggest that the situation is more intricate. For example, Italian DAV expressing meteorological conditions (piovere “rain”, nevicare “snow”, gelare “freeze”) have neither a patient nor an agent, and they oscillate between A and E at any level of the language “without appreciable difference of meaning” (Serianni 1991: 394). “A” would be used in the presence of activity denoting verbs (which are characteristically atelic or not featuring an end point), whereas E would be used when the event is telic (i.e., featuring an end point). Although this contrast explains the difference in the use of A and E with a verb similar to correre “run” (Salvi & Vanelli 2004: 50), it does not explain the choice of AUX in verbs such as cedere “give in” or inciampare “stumble”. Last, some reference grammars of modern Italian suggest that the choice of AUX correlates with the type of syntactic configuration a DAV appears in, regardless of its meaning. In principle, AUX E is used with unaccusative, pronominal, and reflexive verbs, whereas A is used with transitive and unergative verbs. Authors of reference grammars disagree on the exact number of DAVs. For example, unlike Maiden and Robustelli (2000), Schwarze (2009: 154) excludes aumentare “augment, increase”, diminuire “diminish”, crollare “collapse”, and gelare “freeze” from this class, and stresses that some verbs, such as cominciare “begin” and vivere “live”, may take either A or E without apparent reason. Rohlfs (1969: 120) reported that in Tuscan varieties, cessare “cease”, girare “turn”, nuocere “harm”, and tacere “keep quiet” take only A, whereas correre “run” takes only E.

2 Purpose, Delimitations, and Structure of the Article

This article focuses on DAVs in contemporary standard Italian.[3] Its purpose is to identify which factors may influence native speakers’ choice between AUX E and A in a subset of Italian DAVs. The article presents an offline experiment where potentially confounding variables in the choice of either AUXs are controlled. The factors tested in the experiment are what the literature on the unaccusative versus unergative split identifies as a driving AUX selection in DAVs. Due to space limitations, three important aspects were excluded from the analysis. The first is the historic evolution, hierarchy, and interaction of factors that might have an impact on the AUX selection and DAV distribution in the transition from Latin to Romance varieties. Such features include, for example, the consequences of the loss of the Nominative versus Accusative opposition in declension, the loss of the active versus passive opposition in voice, and the reappearance and spread of the active versus inactive opposition in the Romance conjugation starting roughly from the VII century (Pinkster 1987; Loporcaro 2014; Cennamo 2008). The second exclusion concerns AUX distribution across Italian dialects (Loporcaro 2015; D’Alessandro 2017). Such features comprehend, for example, triple auxiliation in Italian dialects and person-driven AUX selection. The third excluded aspect concerns cross-linguistic comparison of auxiliation systems, especially among Romance and Germanic languages (Aranovich 2007; Kailuweit & Rosemeyer 2015). The structure of the article is as follows. In Sections 3 and 4, the theoretical framework is described. In Section 5, the research questions (RQs) are outlined, and in Section 6, the study is described. In Section 7, the results are reported, and in Section 8, they are discussed in light of the RQs and theoretical background.

3 Background

DAVs have been mentioned in syntactic and semantic theories dealing with AUX selection as a diagnostic of unaccusativity since the late 1970s (for a review, see Kailuweit & Rosemeyer, 2015). According to such theories, the choice of either A or E involves both “predicate-level properties” and “clause-level properties” (McFadden 2007, p. 675–682). The former are determined by the interplay of the verb lexical semantics (namely, the telic vs. atelic contrasts), argument theta-roles (namely, the variable degree of agentivity of the grammatical subject), and argument structure, meant as the mapping of event participants onto external or internal argument positions in the verb phrase (VP) (or within the vP shells, e.g., Travis 2010). Also, the (un)quantified nature of the object NP and the role of verb adjuncts—all the elements telicizing, delimiting, specifying the manner in which events take place, or their duration—below the VP are taken into account among predicate-level properties. By contrast, the “clause-level properties” affecting AUX selection are those determined by person and the number of the grammatical subject, tense (past vs. present), mood, and grammatical aspect (perfective vs. imperfective/continuous). Due to space limitations, in this brief review of the literature, only predicate-level properties and adjuncts will be considered, with the aim of identifying the ones that lend themselves to be factorized in an experiment, testing how a subset of educated Italian native speakers living in Northern Italy represent DAVs.

3.1 Early Syntactic Accounts of DAVs

DAVs are mentioned in early syntactic accounts of the unaccusative versus unergative split within the framework of relational grammar (Perlmutter 1978; 1989; Rosen 1987; 1997). On such accounts, AUX selection depends on the nature of the syntactic relation holding between the verb and its only argument in deep and surface representations (“strata”, in Perlmutter’s terms). If the sole argument, which in the initial stratum holds subject-relation with the verb, maintains such relation also in the final stratum, the verb is unergative and the AUX will likely be A. If the original relation of the subject in the initial stratum is instead an object-relation (the subject shares some features with the direct object of transitive verbs), the verb is unaccusative and the AUX will likely be E. Perlmutter (1978; 1989 also proposed that the agenthood or patienthood of the grammatical subject correlates with the subject-relation typical of unergatives and the initial object-relation typical of unaccusatives, respectively. In this framework, DAVs represent a peculiar case because their sole argument can entertain either a subject-relation or an object-relation with the verb. The kind of relation and the AUX that a DAV may take is specified at a lexical level, that is, it depends not only on agenthood, but also on the verb meaning (Loporcaro, 2006: 312). Perlmutter (1978) claimed, in fact, that verb meaning is a reliable cue of its syntactic behavior and predicted that cross-linguistic verbs with equivalent meanings will behave similarly with respect to the intransitive split. Verbs describing intentional or volitional acts (e.g., “work”, “play”, “speak”, “talk”, “ski”, “swim”, “walk”, “cry”, “laugh”, and “dance”) and certain involuntary bodily process (e.g., “cough” and “sneeze”) are likely unergative. By contrast, verbs of existence and happening, verbs meaning nonvoluntary emission of stimuli, and verbs expressing durative events and states are likely unaccusative. Based on these premises, one can assume that native speakers of Italian, even when selecting between A and E in DAVs, will probably look at both classes of meaning and agenthood/patienthood of the subject. Nevertheless, there exist many Italian DAVs in which changes in verb meaning and change in subject agenthood/patienthood are clearly distinct. For example, in sentences (3), the thematic role of the grammatical subject remains identical despite the change in AUX and, consequently, of the verb meaning, whereas in sentence (4) the opposite obtains (in sentence 4b, “castaway” is a theme, that is, the entity who happened to survive).

L’atleta ha/è corso alle olimpiadi
The athlete have/be.aux.3s ran.PastP
‘The athlete competed at the Olympic games’
‘The athlete ran to where the O. took place’
Il naufrago ha/è sopravvissuto in mezzo ai pericoli
The castaway have/be.aux.3s survived.PastP
‘The castaway [theta-role=agent] survived in the midst of danger’
‘The castaway [theta-role=theme] survived in the midst of danger’

Different authors have highlighted the existence of discrepancies between the verb syntactic behavior and its meaning (the so-called “unaccusativity mismatches”, see Alexiadou et al., 2004). However, when reexamining AUX distribution, both within and across languages, Rosen (1987) and Perlmutter (1989) had already recognized that classes of meaning can reliably predict neither syntactic behavior of verbs nor AUX selection in DAVs. In the syntactic approach proposed by Burzio (1986)—in which the subject versus object relation proposed by Perlmutter is substituted by a binding relation between the subject and a nominal linked to (or governed by) the verb before and after A-movement—DAVs are never mentioned, nor does Burzio’s analysis suggest any impact of verb classes on AUX selection.

To sum up, in early syntactic approaches to AUX selection, there is no clear indication of which cues native speakers would rely on when choosing between A and E in DAVs. Perhaps, both “classes of meaning” and “agenthood/patienthood”—alone or in combination—are too coarse-grained for being factorized, and, for this reason, they will not be considered in the current experiment. The syntactic role of the grammatical subject is instead relevant for AUX selection in DAVs, and it will be taken into account, especially in its interplay with verb semantics (Section 3.3).

3.2 Event Structure Account

In sentences (3) and (4) seen above, the presence of either AUX A or E correlates with two different events in which the subject either performs two different actions (sentence 3) or shows a different degree of control on the action (sentence 4). Some syntactic approaches proposed that the unaccusative/unergative split and AUX selection are decided beyond the vebr-phrase (VP), at the level of event structure. In many languages, different verbs—or even the same verb, in the case of DAVs—can enter different syntactic configurations (e.g., reflexive, pronominal, transitive, unaccusative, and unergative) and display different AUXs because the aspectual (±telicity, ±duration) and thematic (agent vs. theme-patient) of the object NP and its(in)definiteness—all potentially licensed by the verb meaning—can be weighed and hierarchized in different ways, depending on the nature of the event, as shown in sentences (3) and (4). The articulated structure that brings about such differences in the event interpretation is a functional projection called “event phrase” (EP). The EP Spec position is where the speaker “measures the event”, that is, they compute whether the event is inherently telic or atelic (McClure 1993). Inherent (a) telicity is purely lexical, that is, noncompositional (see below). Beneath the EP maximal projection, there lies the aspect phrase (AspP), which hosts the VP as its complement (Borer and Alexiadou Anagnostopoulou, 2004, p. 294). The AspP Spec position is where speakers compute whether or not the direct object determiner phrase (DO-DP)—if the verb is transitive—is ±definite by checking quantifiers, (in)definite articles, and accusative case (van Hout, 2004). The definiteness value of the DO-DP adds up to the inherent telicity feature computed higher up in the tree. AspP hosts the VP, which is the layer where the speaker computes the internal versus external nature of the verb sole argument (the grammatical subject) and also its theta-role (agent vs. theme or patient). Given that the EP is higher than the VP, event structure is accessed before argument structure during sentence comprehension (Folli & Harley 2005). Underneath the VP, various XPs adjuncts are hosted, where the telic endpoint, the manner, and the duration of the event are expressed compositionally by means of past participle (PP), adverbs of duration, etc. (for a different account of AspP, i.e., hosted within the vP shell structure, see Travis 1994; 2010). Since all the predicate-level properties described above are encoded in syntax, speakers are expected to check the presence and the structural positions of all relevant language features (e.g., bare plural DPs, quantified object DPs, PP, adverb phrase (AdvP), etc.), which in turn determine the interpretation of the event and the AUX selection.

Some among the EP features described above can be easily factorized in an experiment; others probably cannot. As an example of the latter, it might be difficult to ascertain Italian native speakers’ sensitivity to inherent telicity or atelicity of a predicate at EP Spec position because most predicates per se in Italian—as in other languages—may have multiple actional interpretations (Lenci & Zarcone 2009). As to the factorizable EP features, an experiment can be designed in such a way that participants’ choices between AUX A and E may be influenced by compositional aspect (at XP adjuncts level), that is, by the presence of telic or atelic completions. For example, sentences (3b) and (4b) are derived from sentences (3) and (4) by adding an atelic (“in the midst of danger”) or telic (“until he was rescued”) XP completion, respectively.

L’atleta ha/è corso alle olimpiadi/fino al traguardo
‘The athlete ran at the Olympic games/towards the finish line’
Il naufrago ha/è sopravvissuto in mezzo ai pericoli/fino al salvataggio
‘The castaway survived in the midst of danger/until he was rescued’

The EP approach to AUX selection has theoretical advantages, but it has also left some open issues. On the one hand, it introduced a principle of syntactic flexibility which accounts for certain classes of alternating verbs (including many Italian DAVs). On the other hand, it does not explain why some verbs—including some DAV—do not freely exhibit the same alternation patterns (Folli & Harley 2005, p. 95). This apparent incongruence was dealt with upfront by the proponents of semantic approaches.

3.3 Lexical Semantics Account of DAV

The lexical semantics approach claims that AUX selection in DAV depends solely on the lexical properties of the verb. Foley & Van Valin (1984) and van Valin (1990)—based on Dowty’s (1979) semantic decomposition approach—proposed that AUX selection depends on the lexical aspect—namely, the telic versus atelic contrast—and the degree of agenthood of the grammatical subject expressed in a hierarchy of decreasing values from “actor” (highest agentivity) to “undergoer” (least agentivity). These factors interact: telic and atelic verbs tend to feature different degrees of agenthood of their grammatical subjects. Namely, telic intransitive verbs selecting E tend to have subject bearing the semantic role of theme. On the contrary, verbs selecting A tend to have subject bearing an unmarked actor (it occupies a leftward position in van Valin’s (1990) protoroles hierarchy) (Dowty 1991: 606; van Valin 1990: 233, 256).

Let us return to the split auxiliaries of the DAV correre “run” in sentence (3a) and (3b), repeated here as (5) and (6).

L’atleta ha corso alle Olimpiadi
‘The athlete ran at the Olympics’
L’atleta è corso fino al traguardo
‘The athlete ran at the Olympics’

According to the semantic approach described above, in both sentences, the basic meaning of the verb correre “run” and the grammatical subject are identical, but in the [-telic] interpretation featuring AUX A (5), the grammatical subject bears the agent role, whereas in the [+telic] interpretation featuring AUX E (6), the grammatical subject is a theme.

Bentley & Eythórsson (2004) and Bentley (2006) maintain that such a purely semantic characterization of unaccusativity cannot properly account for AUX selection or DAVs. On the one hand, these scholars agree that split intransitivity in Italian is ultimately determined by the contrast between two types of Aktionsart and the actor–undergoer hierarchy, as in van Valin (1990). On the other hand, they recognize that AUX selection in Italian “is primarily concerned with the semantic status of the argument with the privileged syntactic function of subject of the clause, and it is not possible to account for this phenomenon in purely semantic terms” (Bentley 2006, p. 6). What Bentley (2006) points out is that in modern standard Italian—unlike in old Latin—the syntactic subject relation in the so-called “active alignment” has a privileged role in the semantic representation of any predicate, regardless of the theta-role. Put it differently, in any Italian native speaker’s mental representation of verbs, subjects have a privileged role, as opposed to direct and indirect objects. Therefore, AUX selection in DAVs results from the tension between the enduring (syntactic) subject–object opposition and the underlying semantic actor–undergoer opposition, the latter being a relic of the ancient “unaccusative alignment” that continues to feed a dichotomy that characterizes a subset of Italian intransitive verb. The shift from A to E in Italian DAV, such as sopravvivere “survive” in sentence (4b) seen above, would signal that the external argument (the castaway) maintains some characteristics of Italian subjects (e.g., occupying Spec position, bearing inflection (INFL), being Topic), whereas at the same time, showing a lesser degree of agentivity than subjects of transitive verbs typically show.

Bentley’s (2006) proposal that the grammatical subject of a sentence is the locus where the syntax-semantic tension manifests echo insights from Relational Grammar (Section 3.1) and has important implications for an experiment design aimed at testing native speakers’ competence in DAV use. One remaining issue, however, is that the various degrees of agentivity of the subject are not easily factorizable because they are not categorical and clear-cut. A viable solution is proposed in the approach described in the next section.

3.4 AUX Selection Hierarchy Account[4]

Sorace (2000; 2004; 2015 observed that semantic rules are too rigid, and syntactic accounts are too coarse-grained. In the former case, semanticists cannot explain why there exist systematic exceptions to semantic rules linking arguments and theta-roles; in the latter, syntacticians group together verbs that seem to have nothing to do with one another. Sorace remarked that not all intransitive verbs are born equal: some unaccusatives and some unergatives are, respectively, more unaccusative and unergative than others, and unaccusativity and unergativity are gradient concepts. The AUX selection hierarchy (ASH) put forth by Sorace (2000) postulates that strongly (or “core”) unaccusative or unergative verbs lay at the opposite extremes of a gradient or “hierarchy” (Figure 1), whereas verbs that are less unaccusative or unergative (dubbed as “peripheral”) lay in the middle of the gradient, where more factors mix and interplay. The probability of occurrence of either A or E is a function of the verb position in the gradient. Telic verbs expressing the change of location, such as arrivare “arrive” or venire “come”, occupy one pole of the gradient, are impermeable to compositional factors, and invariably select E. Agentive nonmotional atelic processes, such as parlare “talk” or lavorare “work”, occupy the opposite pole of the gradient and invariably select A. Less semantically specified verbs are located in the middle of the gradient: they are the most variable and open to compositionality, and their AUX choice depends on the interplay between telicity and agentivity, as well as kind of verb completion (i.e., adjuncts). It is among these intermediate categories (written in increasingly pale grey fonts in Figure 1) that Italian DAVs are more likely to be found.

Figure 1: 
Gradient of unaccusativity/unergativity: core verbs are in bold; the less semantically specified verbs are the paler the grey (Sorace 2000).

Figure 1:

Gradient of unaccusativity/unergativity: core verbs are in bold; the less semantically specified verbs are the paler the grey (Sorace 2000).

Telicity and agentivity are not equally relevant for all verbs in the gradient. Telicity is relevant for the upper part of the gradient (verbs of change and verbs of state); agentivity is relevant for the lower part of the gradient (controlled processes). Uncertainty in AUX selection infiltrates and increases as soon as one moves away from the extremes, toward the middle. Verbs expressing the change of state, such as fiorire “bloom” select A when the emphasis is on duration, and telicity is shadowed; they select E in the presence of a punctual adverbial, e.g., in un’ora “in 1 hour”. For verbs expressing continuation of a preexisting state, such as durare “last” and sopravvivere “survive”, the agentivity of the subject interplays with telicity, conveyed, for instance, by a time adverbial bounding the duration of the event. For example, Sorace (2000) suggests that Il presidente è durato in carica due anni “The president stayed in office for two years” is more acceptable that La guerra ha durato due anni “The war lasted two years” because presidente is agentive (and animate). Verbs expressing the existence of a state, such as appartenere “belong” or bastare “suffice”, are the most variable in Italian and may select either E or A (e.g., il cibo ha/è bastato “the food sufficed”). Verbs expressing uncontrolled processes are very variable too. They are divided into five categories: (a) anticausative verbs (externally caused events whose causer has been suppressed or demoted), e.g., aumentare “increase”, diminuire “decrease”, avanzare “move forward”, continuare “continue”, migliorare “improve”, and peggiorare “worsen”; (b) verbs strongly affecting the subject, e.g., rabbrividire “shiver”, scarseggiare “be lacking”, and sbandare “swerve, stray”; (c) bodily process, e.g., tossire “cough”, sudare “sweat”, and vomitare “vomit”, where the subject is animate and affected, but not agentive; (d) weather verbs, e.g., piovere “rain”, nevicare “snow”, and grandinare “hail”; and (e) emission verbs, e.g., squillare “ring”, suonare “sound”, risuonare “resound”, and lampeggiare “flash”, where lower animacy counterbalances high subject affectedness. Finally, verbs expressing controlled process encompasses those affecting the subject (e.g., aderire “adhere”, cedere, “give in”, and trionfare “triumph”) and motion verbs (e.g., correre “run”, rotolare “roll", and atterrare “land”).[5]

As pointed out at the end of Section 3.3, the gradient of agentivity of the grammatical subject (considered as the key factor in AUX selection by the semantic approach) is not easily factorizable. The solution found by Vernice & Sorace (2018) is to use instead minimal pairs of sentences featuring animate versus inanimate subjects, as the ±animate dichotomy is a much simpler and more categorizable value than those found in the actor–undergoer hierarchy. Vernice & Sorace (2018, p. 850) appealed to language typology literature which supports the view that—at least with transitive verbs—animate agents are more prototypically agentive than inanimate agents, and thus more accessible. For this reason, Vernice & Sorace (2018), in their experiment, presented participants with inanimate subjects (e.g., “the moped”) and animate subjects (e.g., “the rebel”).

As to the testability of theoretical predictions of the ASH, four among them seem particularly suitable for an experiment on DAV: (a) DAV must be searched among the peripheral verbs of the gradient; (b) compositional aspect (the role of telic vs. atelic adjuncts) is key for AUX selection in DAV; (c) The ± animacy of the subject is more suitable as an experimental factor than its ± agentivity; and (d) The impact of animacy and telicity is selective and should be tested for different verb classes, i.e., controlled processes and verbs of change/state. There are only a few behavioral studies focusing on how native speakers process peripheral and core verbs along the gradient proposed by the ASH. Studies employing an offline magnitude estimation task (Bard et al. 1996) indicate that native speakers have clearer intuitions on the AUX of core verbs when compared with that of the peripheral verbs. Kraš (2010) tested the linguistic intuitions on auxiliaries of 16 highly proficient adult learners of Italian (L1 Croatian) by using a speeded acceptability judgment task. The acceptability score and response time means showed that subjects distinguished between core and peripheral verbs when selecting AUX A or E. In their eye-tracking study, Bard et al. (2010) found that the incorrect AUX caused longer total reading times in sentences with core unaccusative or unergative verbs in comparison with the peripheral verbs. Finally, in a more recent eye-tracking study, Vernice & Sorace (2018) monitored the processing of animacy of intransitive verb arguments and found that inanimate subjects caused longer reading times only for unergative verbs, whereas with unaccusative verbs, argument animacy did not influence reading times.

3.5 Syntactic Alternations Account

Although we have seen that some DAVs such as sopravvivere “survive” retain their core meaning while changing the AUX, the different syntactic configurations in which some DAVs may occur could highlight different event structures, namely, different phases of the same core event (Moens & Stedmann 1995). For example, the DAV bruciare “burn”, when used intransitively, can be either unergative, unaccusative, or pronominal (anticausative), and in combination with different time expressions, can underline duration (7), endpoint (8), and resulting state of the event (9).

Il bosco ha bruciato per due ore
‘The woods burned for 2 h’ (unergative, focusing on event duration)
Il bosco è bruciato fino al mattino seguente
‘The woods burned until the following morning’ (unaccusative, focusing on the endpoint of the event)
Il bosco si è bruciato in pochi minuti
‘It only took a few minutes for the woods to burn down completely’ (pronominal, focusing on time to the resulting state)

Interestingly, unlike the DAV bruciare “burn”, other DAVs such as inciampare “stumble” and volare “fly” have neither a transitive nor a pronominal counterpart. This difference among Italian DAVs must be taken into account in the design of an experiment targeting native speakers’ competence in use of DAVs. It is in fact possible that different syntactic configurations associated with the same verb could trigger a priming effect. The AUX of the prevalent configuration (e.g., the most frequent, similar, or contiguous in meaning) could be the one that comes to speakers’ mind first, regardless of either DAV syntax or semantics. “Resonance” could be a proper name of the effect produced by the coexistence, within a speaker’s competence, of the various configurations in which the same verb enters. Under this respect, the classification of Italian DAVs across syntactic configurations is quite intricate. Jezek (2003) subdivided Italian predicates into 15 categories depending on their unique or multiple (transitive, unergative, unaccusative, and pronominal) alternations. Many of these verbs completely change their meaning depending not only on the AUX, but also on the type of subject. For example, saltare can mean “skip” when transitive (e.g., L’oratore ha saltato i saluti “The speaker skipped the greetings”), or it can mean “jump” when unergative (e.g., L’atleta americana ha saltato per prima “The American athlete jumped first”). The same verb can also mean “go out” when unaccusative (e.g., La luce è saltata “The light went out”). Other DAVs with transitive and/or pronominal counterparts are gelare “freeze”, aumentare “increase”, migliorare “improve”, continuare “continue”, mancare “miss”, fondere “melt”, and fallire “fail”. Resonance can be attractive or repulsive. In the former case, a speaker may choose AUX A because the unergative verb resembles or is semantically contiguous to the transitive verb. In the latter, a speaker might avoid choosing AUX E because it can be confounded with the passive or adjectival use of the verb. Attractive resonance mostly concerns the transitive counterpart of unergative verbs such as continuare, migliorare, and peggiorare. As shown in (10), the transitive (a) and the unergative (b) versions of continuare “continue” are indeed semantically contiguous.

Mario ha continuato la corsa
‘Mario continued the run’
Mario ha continuato a correre
‘Mario kept on running’

The attractive resonance between configurations (a) and (b) may lead speakers to choose AUX A across the board, and it may also contribute to explain why, in our corpus of reference (ItTenTen, see below), continuare is five times more frequent with AUX A than with AUX E. Repulsive resonance may instead occur between a verb similar to cedere “give in” and AUX E, because E is also the AUX used in the passive (meaning “being given up”), which in Italian corpora is way more frequent than the intransitive. In this paper, only the effect of attractive resonance is tested.

3.6 Frequency Account of DAVs

The frequency at which DAVs feature A or E in the input may determine native speakers’ choices between AUX in real-time processing, by interacting, competing, and even overriding syntax and semantics. Yet, frequency has not been dealt with adequately in the literature. For example, Vernice and Sorace (2018) normed verbs and sentences for frequency by checking verb frequency on the “Corpus e Lessico di Frequenza dell’Italiano Scritto” (CoLFIS).[6] In their analysis, however, frequency in the input was not calculated separately for each AUX + PP combination (e.g., the frequency of ha corso vs. the frequency of è corso), but only by categories that are relevant for the ASH (unaccusative vs. unergative verbs and core vs. peripheral verbs). A more accurate analysis could have shown that some AUX + PP combinations occur at negligible frequency in the input (if at all). For instance, important categories of verbs, such as those expressing continuation of a preexisting state or existence of a state, in which DAVs are extremely common, never occur in authoritative corpora of spoken and written contemporary Italian. As noted by Sorace (2000; 2004; 2015, strings featuring AUX A + PPs durato, rimasto, bastato, appartenuto, and intervenuto are not attested in Lessico dell’Italiano Parlato,[7] and in the Corpus di Italiano Scritto (CORIS),[8] the unergative and unaccusative versions appear at very different frequencies. For example, unergative ha durato “it lasted” has two occurrences, whereas unaccusative è durato has about 1000 occurrences; unergative ha bastato “it sufficed” has zero occurences, whereas unaccusative è bastato has about 1200 occurrences; and unergative ha appartenuto “it belonged” has four, whereas its unaccusative counterpart è appartenuto has about 100 occurrences. To provide an example from another verb category, the unergative version of scendere “descend, go down”, featured in (11), never occurs among the 150 million words of the CORIS, as observed by Sorace (2004: 259).

La popolarità del governo ha sceso notevolmente
‘The government popularity decreased dramatically’

One may object that—although basically no one in Italy uses them—at least some of the DAVs proposed by Sorace could be virtually represented in a native speaker’s mental competence.

4 To Sum Up: Testable Key Factors that Drive AUX Selection in DAVs

Some key factors driving AUX selection in DAVs lend themselves to be tested in an experiment, whereas others do not. Among the testable traits, the following ones stem directly from the theoretical accounts described in Section 3:

  1. Bentley’s (2006) and early syntactic approaches in the Relational Grammar framework suggested that the peculiar syntactic status of the grammatical subject plays a pivotal role in AUX selection in DAVs, perhaps regardless of its semantic role;

  2. The event structure syntactic approach suggested that it is in position below the VP, where XP adjuncts are located, that compositional telicity is computed, and the aspectual flavor and the interpretation of a sentence are decided;

  3. The ASH suggested that (a) DAVs are to be found among the “peripheral verbs” in the gradient; (b) animacy is more easily testable than values scattered on the agentivity gradient (as proposed in the semantic approach); and (c) animacy and telicity might affect different verbs, namely, verbs expressing a state/change of state versus an uncontrolled nonmotional process;

  4. The existence of alternating DAVs in Italian (Jezek 2003) suggested that there might exist a resonance effect among different syntactic configurations a verb can enter in Italian. Resonance may affect native speakers’ AUX choice in DAVs;

  5. The analysis of Italian spoken and written corpora suggested that frequency and association may impact heavily on participants’ choice of either AUXs and even overwrite syntactic and semantic factors.

5 Research Questions

The key factors derived from the theoretical background illustrated in the previous section motivated four RQs this study aims to address:

  1. RQ1: Did the presence of a ±telic completion influence participants’ choices in the case of stimuli featuring a change of state and stative DAVs?

  2. RQ2: Did the presence of a ±animate subject influence participants’ choices in the case of stimuli featuring nonmotional process DAVs?

  3. RQ3: Did the (variously modeled) frequency of AUX + PP combinations in the input influence participants’ choices?

  4. RQ4: Did resonance (i.e., the presence of transitive or pronominal counterparts of DAVs) influence participants’ choices among competing stimuli?

6 Study

6.1 Materials: DAV Selection and Manipulation of the Stimuli

For this experiment, 32 DAVs were selected from the peripheral verb list compiled by Sorace (2000; 2004, Section 3.4) and from the verbs indicated by Jezek (2003: 194–200, Section 3.5) as taking both AUXs in contemporary Italian. The 32 DAVs were divided into two pools: “alpha” and “beta”. Pool alpha contained 16 state or change of state verbs, for which ±telicity is held to be relevant in the literature. Pool beta contained 16 verbs of nonmotional processes, for which subject animacy is held to be relevant. These DAVs were entered into a 2 × 2 factorial design with ±telicity (alpha pool) and ±animacy (beta pool) as the independent variables, in order to generate four experimental sentences for each DAV. For example, each DAV from pool alpha generated four sentences, one featuring AUX A + telic completion, another featuring AUX A + atelic completion, a third one featuring AUX E + telic completion, and a fourth one featuring AUX E + atelic completion. The levels of the independent variables were manipulated as follows. For the 16 sentences featuring DAVs from pool alpha, telic completion was obtained by adding punctual time adverbs or adjuncts that specify the endpoint of the event, either in time or space (e.g., “in a few days”, “until midday”, and “to the edge of the goal are”). Atelic completions, on the other hand, were obtained by adding durative time adverbs (e.g., “for two years”) or by specifying the circumstances of the event (e.g., “in the midst of the enemies”, “in spite of the cold”, “completely alone”, and “spontaneously”). The stimuli were built as follows; when telicity was the variable (i.e., in pool alpha), the same grammatical subject (underlined in (12)) had to be compatible with both telic and atelic sentence completions, whereas when animacy was the variable (i.e., in pool beta), the sentence completion (underlined in (13)) had to be compatible with both animate and inanimate subjects.

L’impero romano ha/è sorto in mezzo ai nemici/in mezzo secolo
‘The Roman Empire rose in the midst of enemies/in half a century’
L’atleta americano/l’accordo sui salari ha/è saltato per primo
‘The American athlete/the salary agreement jumped/fell through first’

The number of words in the experimental sentences ranged from 6 to 11. In order to keep the PP ending (i.e., -o) identical across AUX, sentence subjects were all masculine and singular.[9] Tables 1 and 2 list, respectively, the experimental sentences of pool alpha and pool beta, with their English translation.

Table 1:

Sixty-four experimental sentences featuring pool alpha DAV (i.e., change of state and stative verbs).

Verbs of state and change of state DAVs Experimental sentences
sorgere “rise” (1) L’impero romano ha/è sorto in mezzo ai nemici/in circa due secoli
“The Roman empire rose in the midst of enemies/in half a century”
fiorire “bloom” (2) L’albero ha/è fiorito malgrado il freddo/in pochi giorni
“The tree bloomed despite the cold/in a matter of a few days”
germogliare “sprout” (3) L’albero ha/è germogliato in pieno inverno/in poco tempo
“The tree sprouted in the midst of winter/in a short time”
marcire “rot” (4) Il legno del tetto ha/è marcito nonostante la ristrutturazione/in pochi anni
“The wood of the roof rotted despite having been treated/in a matter of a few years”
sopravvivere “survive” (5) Il naufrago ha/è sopravvissuto in mezzo ai pericoli/fino al salvataggio
“The castaway survived in the midst of danger/until he was rescued”
durare “last” (6) L’assedio ha/è durato per due anni/fino al 1924
“The siege lasted two years/until 1924”
appartenere “belong” (7) L’oro ha/è appartenuto agli eredi per generazioni/fino al 1850
“The gold has belonged to the heirs for generations/belonged to the heirs until 1850”
aumentare “increase” (8) Il prezzo del petrolio ha/è aumentato senza una ragione/fino a 10 euro al barile
“The price of oil increased without reason/to 10 euros per barrel”
diminuire “decrease” (9) Il dolore ha/è diminuito spontaneamente/fino a sparire
“The pain decreased naturally/until it was imperceivable”
avanzare “move forward” (10) L’attaccante ha/è avanzato senza trovare ostacoli/fino al limite dell’area
“The striker moved forward without opposition/to the edge of the goal area”
continuare “continue” (11) Il cantante ha/è continuato malgrado i fischi/fino alla fine del concerto
“The singer kept on singing despite heckling/until the end of the concert”
ingrassare “gain weight” (12) L’attore ha/è ingrassato per il nuovo film/fino a diventare irriconoscibile
“The actor gained weight for the new movie/until he became unrecognizable”
migliorare “improve” (13) Il suo impegno scolastico ha/è migliorato notevolmente/fino alla fine del semestre
“His performance in school improved dramatically/kept on improving until the end of the semester”
peggiorare “worsen” (14) Il tempo ha/è peggiorato costantemente/nel giro di pochi minuti
“The weather worsened consistently/in a matter of a few minutes”
resistere “resist” (15) Il campanile ha/è resistito molti secoli/fino al bombardamento
“The clock tower has endured throughout the centuries/stood until the bombing”
correre “run” (16) L’atleta ha/è corso alle olimpiadi/fino al traguardo
“The athlete ran at the Olympic games/towards the finish line”

Table 2:

Sixty-four experimental sentences featuring pool beta DAV (i.e., nonmotional and uncontrolled process verbs).

Uncontrolled process DAVs Experimental sentences
funzionare “operate, work” (1) Il discorso ufficiale/il giovane presidente ha/è funzionato da stimolo per la nazione
“The official speech/the young president served as an inspiration for the country”
aderire “adhere, stick” (2) Il rocciatore/l’ornamento ha/è aderito perfettamente alla parete
“The cragsman perfectly clung/the ornament perfectly stuck to the wall”
cedere “concede, give in” (3) Il tetto/il vecchio tiranno alla fine ha/è ceduto sotto il peso degli anni
“The roof finally collapsed due to its age/the old tyrant finally conceded due to his age”
saltare “jump”, “go off”, “fall through” (4) L’atleta americano/l’accordo sui salari ha/è saltato per primo
“The American athlete jumped first/the salary agreement was the first to fall through”
trionfare “triumph, prevail” (5) Il candidato alle elezioni/il nuovo software ha/è trionfato su tutti i competitori
“The campaigner prevailed over his competitors/the new software beat out its competitors”
rotolare “roll” (6) Il soldato dei corpi speciali/il sasso ha/è rotolato lungo il pendio
“The special forces soldier/the stone rolled down the slope”
atterrare “land” (7) Il pilota/l’aereo ha/è atterrato malgrado le pessime condizioni meteo
“The pilot/the plane landed despite the bad weather”
sbandare “swerve” (8) Il SUV/il pilota ha/è sbandato in curva finendo fuori strada
“The SUV/The driver swerved on the curve and went off the road”
tremare “tremble, shake” (9) Durante il bombardamento, l’uomo/l’edificio ha/è tremato vistosamente
“During the bombing, the man/the building was visibly shaking”
tentennare “hesitate”, “vacillate” (10) l’ombrellone/l’ubriaco ha/è tentennato sotto le forti raffiche di vento
“The beach umbrella shuddered/the drunk man vacillated with the gusts of the wind”
circolare “circulate”, “operate” (11) Il denaro falso/il criminale ha/è circolato per anni sotto gli occhi della polizia
“The fake currency circulated/the criminal operated for years undetected by the police”
decollare “take off” (12) Il pilota/l’aereo ha/è decollato in perfetto orario
“The pilot/the plane took off perfectly on time”
convivere “coexist” (13) Il sistema economico/il nuovo sindaco ha/è sempre convissuto con il sistema politico
“The economic system/the new mayor has always coexisted with the political system”
inciampare “stumble” (14) il testimone/il comunicato stampa ha/è inciampato in molte contraddizioni
“The witness stumbled through/the press release was filled with many contradictions”
prevalere “prevail” (15) Lo spirito di squadra/l’atleta alla fine ha/è prevalso sugli avversari
“The team spirit/the athlete prevailed over the other contestants in the end”
brillare “shine, twinkle” (16) Stavolta l’astro/lo studente ha/è brillato meno del solito
“This time the star/the student shone less than usual”

The naturalness of experimental sentences was ensured by 27 Italian native speakers—all graduate students at Italian Universities—prior to the experiment, who were asked to rate the compatibility of each verb in the telic and atelic completion conditions (pool alpha), in the animate and nonanimate conditions (pool beta). They used a five-point scale, ranging from 1 (= “absolutely unnatural”) to 5 (= “absolutely natural”), with intermediate values of “somewhat unnatural”, “somewhat natural”, and “natural”. The average score was 4.6, and between raters’ agreement was always ≥90%, with no significant differences between pools (one-way ANOVA, p-value = 0.23). The degree of reliability of raters’ responses (measured with Cronbach’s α) was always ≥ 0.88 (function alpha from the R package “psych”). The raters proved homogeneous and consistency in their judgments. The naturalness of sentences was also checked via a debriefing questionnaire.

6.2 Coding of Frequency, Association Scores, and Resonance

In this study, the frequency of AUX + PP combinations in DAV was factorized in three ways by using (a) raw frequency scores of each AUX + PP combinations; (b) bidirectional association scores between either AUX and PP, and (c) unidirectional backward transition probabilities[10] (BTPs) between the PP and either AUX A or E. The raw frequency reflects token frequency (number of identical repetitions) of each AUX + PP pair (or, more technically, “collocation”) in the input. The association scores reflect the frequency of cooccurrences of AUX and PP (that is, their association strength) and can be measured via nondirectional (or unidirectional) statistics, such as association tests. The association tests measure the probability that two or more words cooccur with higher-than-average probability in the input. Different PPs behave differently with respect to their association with AUX. For example, the PP aumentato “augmented” “increased” cooccurs with a higher-than-average probability with both AUX E and A, whereas atterrato “landed” and decollato “taken off” are much less frequently associated with either AUXs. In this study, I utilized T-score and LogDice as association scores. The BTP reflects the probability that a word (in our case, the PP) is preceded by another word (in our case, the AUX). The use of BTP rather than “forward transition probabilities" is motivated by the fact that each AUX + PP is a combination of words where the latter (the PP) are far more varied in the input than the former (because there are only two auxiliaries). Since the relative frequency of each component of the AUX + PP collocation is unbalanced, statistical representations of AUX + PP collocations in the native speaker’s mind are likely to originate from BTP (i.e., the probability that a word is preceded by another) (Onnis et al. 2008), rather than from cloze predictability (i.e., the probability that a word is followed by another). In other terms, AUX and PP form a collocation whose head is the second element (Malec 2010: 129), and the direction of headedness is backward. For example, in written Italian, the probability that AUX A is followed by squillato “rung” is approximately 100,000 times lower than the probability that squillato is preceded by A (= 0.42). Therefore, any AUX + PP pair is an asymmetric collocation whose predictive element (in our case, squillato) follows the AUX instead of preceding it. The BTPs are calculated by entering frequency values in a contingency table (Levshina 2015). The normalized (per million) raw frequency, association scores, and BTPs of the 32 DAVs selected for the study were calculated with itTenTen, which, to date, is by far the largest Italian corpus. The itTenTen16[11] is a 4.9-billion-word web corpus made up of texts collected from the Internet (via SpiderLing[12] between May and August 2016). The corpus is a part of the TenTen corpus family, a set of the web corpora built using the same method (Jakubíček et al. 2013).

The raw frequency of DAV was calculated with function “concordance” of SketchEngine. The corpus query language (CQL) was used to refine the search and to filter out transitives, absolute participle phrases, and passives.[13] , [14] The BTPs between PP and AUX were calculated with the formula in (14), which express the probability (p)—ranging from 0 to 1—that a given PP is preceded by either AUX A or E.

( P ( A U X | P P ) ) = f r e q A U X + P P f r e q P P

The calculation of BTPs also includes transitive (taking A) and pronominal (taking E) alternations.

As previously mentioned, LogDice and T-score were considered as association scores in this study. The LogDice formula expresses the typicality of a collocation by separately comparing the frequency of the node (AUX) and the frequency of the collocate (PP) with the frequency of the whole AUX + PP collocation (i.e., how frequently the node and the collocate cooccur). LogDice is not influenced by the corpus size.

By contrast, the T-score expresses the probability that AUX and PP do not cooccur randomly and is calculated with the formula in (15), where “f(node, collocate)” indicates the frequency of the cooccurrence frequency (f) of the PP (node) and the lemma avere “have” or essere “be” of the AUX (collocate). Finally, “DimCorpus” is the corpus size expressed in tokens.

T S c o r e = ( f ( n o d e , c o l l o c a t e ) f ( n o d e ) f ( c o l l o c a t e ) D i m C o r p u s ) f ( n o d e , c o l l o c a t e )

The formula shows that, unlike LogDice, the T-score is highly influenced by the raw frequency of the AUX + PP collocation and therefore, compared to LogDice, is less indicative of the statistical significance of the cooccurrence between the node and the collocate. SketchEngine can calculate the association scores between the node (PP) and collocates (AUX) through the function “collocation”.[15]

Finally, the number of syntactic alternations—transitive and pronominal—for each DAV was derived from the analysis of Italian dictionaries following Jezek (2003). Table 3 reports the frequency and association scores, BTPs, and alternations for each DAV in pools alpha and beta.

Table 3:

Frequency, association scores, BTP, and syntactic alternations (tran = transitive; pron = pronominal) of the DAVs used in the experiment.

Pool DAV freqA freqE LogDiceA LogDiceE TscoreA TscoreE btpA btpE tran pron
Α sorgere “rise” 0.001 4.83 −3.58 2.85 −28 157 0 0.4 no no
Α fiorire “bloom” 0.03 0.5 −2.83 −0.36 8.62 49.96 0.04 0.64 no no
Α germogliare “sprout” 0.03 0.12 −2.9 −2.39 12.80 24.69 0.08 0.27 no no
Α marcire “rot” 0.003 0.05 −6.49 −3.63 0.29 15.12 0.008 0.16 no yes
Α sopravvivere “survive” 0.03 2.59 −3.24 2 −3.78 117.91 0.006 0.61 no no
Α durare “last” 0.08 4.97 −1.69 2.96 8.38 161 0.006 0.38 no no
Α appartenere “belong” 0.07 0.63 −1.86 −0.02 12.56 54.10 0.021 0.19 no no
Α aumentare “increase” 1.20 13.44 3.49 4.40 120 268 0.10 0.48 yes no
Α diminuire “decrease” 0.22 4.92 1.22 1.53 56.86 99.78 0.05 0.53 yes no
Α avanzare “move forward” 1.14 2.48 2.69 1.96 77.79 76.45 0.037 0.05 yes no
Α continuare “continue” 10.5 2.16 5.57 1.76 265 88.22 0.55 0.09 yes no
Α ingrassare “gain weight” 0.01 0.33 −2.89 −0.97 11.44 41.32 0.03 0.38 yes yes
Α migliorare “improve” 0.63 4.23 2.95 2.94 3.98 3.97 0.13 0.32 yes yes
Α peggiorare “worsen” 0.10 1.26 0.52 0.98 43.52 81.50 0.11 0.41 yes yes
Α resistere “resist” 2.82 0.06 3.44 −3.49 127.39 −2.84 0.90 0.01 yes no
Α correre “run” 2.91 1.67 2.96 −2.3 107.91 21.36 0.18 0.10 yes no
Β funzionare “operate, work” 4.75 0.06 4.19 −3.48 165 −19 0.84 0.01 no no
Β aderire “adhere, stick” 10.48 0.02 5.33 −4.76 245 −111 0.88 0.001 no no
Β cedere “concede, give in” 3.77 0.01 3.89 1.62 147 95 0.36 0.19 yes no
Β saltare “jump”, “go off”, “fall through” 0.9 2.81 2.19 2.14 79.36 121.48 0.15 0.39 yes no
Β trionfare “triumph, prevail” 0.99 0.005 1.93 −7.17 75.44 −21 0.88 0.003 no no
Β rotolare “roll” 0.02 0.13 −3.38 −2.34 10.80 25.74 0.06 0.39 yes no
Β atterrareland 0.09 0.91 −1.38 0.51 20.05 69.66 0.04 0.45 no no
Β sbandare “swerve” 0.12 0.08 −1.21 −2.94 25.09 20.20 0.35 0.25 no no
Β tremare “tremble, shake” 0.25 0.01 −0.08 −5.46 37.50 3.89 0.61 0.03 no no
Β tentennare “hesitate”, “vacillate” 0.06 0 −2.23 −10.28 17.81 −3.68 0.72 0.006 no no
Β circolare “circulate”, “operate” 0.15 0.55 −0.79 −0.22 271.5 53.37 0.10 0.36 no no
Β decollare “take off” 0.02 0.43 −3.58 −0.57 6.65 47.63 0.02 0.40 no no
Β convivere “coexist” 0.34 0.03 0.61 −4.46 47.64 8.68 0.70 0.05 no no
Β inciampare “stumble” 0.06 0.25 −1.87 −1.17 19.46 39.41 0.15 0.59 no yes
Β prevalere “prevail” 1.52 0.48 2.65 −0.28 96.60 49.36 0.55 0.17 no no
Β brillare “shine, twinkle” 0.52 0.03 1.43 −4.36 63.48 5.35 0.62 0.03 no no

As shown in Table 3, raw frequency, association scores, and BTPs may not correlate. For example, DAVs aumentare “increase” and diminuire “decrease” with aux A and E have incomparable raw frequency, similar T-scores, almost identical LogDice, and incomparable BTPs. By contrast, with the verb avanzare “move forward”, AUX A and E display similar values in all respects. The inconsistencies between scores may depend on the formulae used to determine association scores (that is, whether or not such formulae take the corpus size as a factor). By contrast, inconsistencies between association scores and BTPs depend on the fact that while association scores reflect mutual attraction (i.e., how likely it is for AUX and PP to cooccur), BTP reflects unidirectional attraction (how likely it is for the PP to be preceded by either AUX). The latter probability depends on PP contingency: if in the input a given PP frequently occurs in other constructions (e.g., passives), its reliability to the AUX + PP construction decreases.

AUX distribution in the sampled DAV was well-balanced. AUX E is overall slightly more frequent than AUX A in our reference corpus (respectively, 1.64 vs. 1.34 occurrences per million tokens on average). In our sample, AUX A and E have identical mean BTP (A = 0.28, E = 0.27), similar mean Tscore (A = 0.66, E = 0.56), and mean LogDice (A = 0.02, E = −0.9). With respect to their distribution, AUX E is on average significantly more frequent in pool alpha—where telicity is manipulated—than in pool beta—where animacy of the subject is manipulated—(E = 2.83, A = 0.38), whereas the opposite holds in pool beta (A = 1.56, E = 1.12), albeit such difference does not reach statistical significance.[16] Association scores and BTPs consistently follow the same pattern: values are higher for AUX E in pool alpha and for AUX A in pool beta. This warns against the risk of confounding frequency and syntactic-semantic factors in the analysis. As a matter of fact, speakers may well choose AUX A with pool beta DAV because of animacy but also because, with such verbs, A is overall more frequent than AUX E in the input, regardless of semantics.

6.3 Sentence Pairing

As mentioned, the 32 DAVs selected generated 128 (32 × 4) experimental sentences, which were eventually arranged in identical pairs, differing only with respect to the AUX. There were 64 sentence pairs in total, 32 from pool alpha and 32 from pool beta. Sentence pairs from pool alpha opposed AUX E and A, once in the [+telic] completion condition and once in the [-telic] completion condition, as shown in sentence (16) and Table 4.

Table 4:

Creation of a stimulus pair featuring a pool alpha DAV.

Pair Subject AUX DAV Telic completion Atelic completion
1 L’impero Romano haè Sorto in circa due secoli
2 haè in mezzo ai nemici
L’impero romano ha/è sorto in mezzo ai nemici/in circa due secoli
‘The Roman Empire rose in the midst of enemies/in about two centuries’

Sentence pairs from pool beta opposed AUX E and A, once in the [+animate] subject condition and once in the [-animate] subject condition, as shown in sentence (17) and Table 5.

Table 5:

Creation of a stimulus pair featuring a pool beta DAV.

Pair Animate subject Inanimate subject AUX DAV Completion
1 Il pilota L’aereo Haè decollato in perfetto orario
2 Haè
Il pilota/L’aereo ha/è decollato in perfetto orario
‘The pilot/the airplane took off perfectly in time’

In order to avoid that the same participant saw the same DAV featuring both AUX under both conditions (±telic or ±animate), sentence pairs featuring the same DAV were arranged into two balanced presentation lists, each containing the same number of predicates from either pool under the same conditions. Participants were randomly assigned to either presentation list. Thirty-five subjects were presented with list 1, and 27 with list 2. Sentence pairs were presented in a pseudorandomized order (different for each list) in order to avoid proximity effect. Sixteen fillers (targeting gerund use, word order, and cliticization) were added to each list. Each list contained 32 target sentence pairs and 16 fillers, for a total of (64 + 16) 80 items.

6.4 Participants

Sixty-two subjects (range 19–28, mean age 23.4, SD = 1.28) took part in the experiment. All subjects were undergraduate and graduate students at different universities in Northern Italy. All subjects have been living in Northern Italy for at least 5 years at the time of the experiment. A short pretest questionnaire indicated that participants’ mother tongue and language of primary exposure (i.e., the language of instruction and everyday communication) was standard Italian, although the majority of them (63%) stated that they had grown up in a dialect-speaking environment (see below). Table 6 and Figure 2 report the dialects or regional variety participants indicated they had been exposed to before moving to Northern Italy.[17]

Table 6:

Dialects (or regional varieties) spoken in the participants’ households.

Dialect spoken in the family Central Italy None Northern Italy Sardinian Southern Italy
% 16 37 24 5 18
Figure 2: 
Breakdown of the dialects (or regional varieties) in the participants’ families.

Figure 2:

Breakdown of the dialects (or regional varieties) in the participants’ families.

In order to ascertain which kind of competence (passive or active) of the dialects or regional variety participants had at the time of the experiment, a backup recall questionnaire based on a reduced version of Grassi et al. (1997, p. 161–269) was administered to all participants 1 year after the test via Google survey. All participants except three responded to the backup questionnaire, consisting the following four questions: (1) “Which language do your parents use when they speak to each other?”; (2) “In what language or dialect did you learn to speak?”; (3) “Do you understand the dialect that spoken in your hometown?”; and (4) “Can you speak it?”. The analysis of the answers showed that all participants—although 63% of them had grown up in a dialect speaking environment—were monolingual speakers of standard Italian, with exclusive use of Italian in the family since birth. Therefore “exposure to a dialect speaking environment” should not be interpreted in terms of capacity of understanding and speaking a dialect. Yet, only 13 among them (21%) admittedly had a limited—exclusively passive—competence of local dialect or regional variety, especially those of Northern Italy. Based on the high homogeneity of outcomes of this posttest recall questionnaire, it was decided to exclude the region of birth from the independent variables considered in the current study.

6.5 Method

In this prompted forced-choice test, participants saw one pair of sentences at a time on a Powerpoint slide. In each pair, one sentence was labeled “A”, and the other “B”. Participants had to choose between sentence A and sentence B on a Google Drive online questionnaire, whereas the sentences were still on the screen. The sentences were presented on a PC monitor following a fixation cross. Participants had 9 s to choose each sentence before an acoustic signal occurred and a fixation cross-signaling the next pair was about to come appeared. Participants could not go back once they had made their choice. Nonrated pairs (1% of the total) were excluded from the analysis. A warm-up trial allowed participants to familiarize with the timing. The experimental session lasted approximately 20 min in total, including preliminary instructions and warm-up.

The experiment adopted a within-subject and within-item design. All participants experienced every condition (telic vs. atelic, animate vs. inanimate), and each item was presented equally often in each experimental condition. Participants were randomly assigned to one sentence lists, balanced across experimental conditions. The order of sentence within each list was pseudorandomized to avoid adjacency effects. Task instructions were as follows: “For each pair of sentence A and B, we ask you to tell us which one sounds more natural in Italian to you.” The meaning of “naturalness” was discussed at length with participants before the trial. It was clarified that both sentences were acceptable in Italian and that the choice of either one was a matter of personal preference. The terms “grammatical” and “grammaticality” were accurately avoided in both oral explanation and written instructions. The experimental setup was meant to assure that (a) each participant saw multiple items per condition; (b) items in each pair formed a minimal pair; (c) lexical repetition, adjacency effects, and familiarization to the task were avoided; and (d) the different experimental conditions were disguised as much as possible. The test was administered simultaneously to all participants through the Web (Zoom) due to the restrictions imposed by Covid-19 protocols. Given the uncertainty due to the current Covid-19 outbreak, it was decided to adopt the Protocol for Remote Testing for the Second Language Psycholinguistics Lab of Indiana University.[18] Targeted recruitment was preferred (via professional contacts, as well as “word of mouth”). The task was limited in scope and time; Internet stability issues were acknowledged and contained; the environmental noise was rated by participants and circumscribed, and the use of headphones was suggested. A final debrief was provided; all responses were collected online and kept anonymous.

7 Results

7.1 Participants’ Responses by DAV and by Pool

Table 7 reports participants’ choices between competing stimuli—containing either AUX A or E.

Table 7:

Participants’ choices between stimuli featuring either A or E for each DAV (in percentage).

DAV % A % E DAV % A % E DAV % A % E
aderire “adhere, stick” 80 20 appartenere 0 100 atterrare “land” 5 95
aumentare “increase” 1.6 98.4 avanzare “move forward” 35.5 64.5 brillare “shine, twinkle” 83.8 16.2
cedere “concede, give in” 74.2 25.8 circolare “circulate, flow” 54.8 45.2 continuare “continue” 100 0
convivere “coexist” 79 21 correre “run” 92 8 decollare “take off” 24.2 75.8
diminuire “decrease” 0 100 durare “last” 0 100 fiorire “bloom” 38.7 61.3
funzionare “operate, work” 98.4 1.6 germogliare “sprout” 48.4 51.6 inciampare “stumble” 40.3 59.7
ingrassare “gain weight” 0 100 marcire “rot” 0 100 migliorare “improve” 0 100
peggiorare 8 92 prevalere “prevail” 80.6 19.4 resistere “resist” 80.6 19.4
Rotolare 11.3 88.7 saltare “jump”, “go off”, “fall through” 54.8 45.2 sbandare “swerve” 77.4 22.6
sopravvivere 0 100 sorgere “rise” 1.6 98.4 tentennare “hesitate”, “vacillate” 87 13
Tremare 93.6 6.4 trionfare “triumph, prevail” 100 0
“tremble, shake”

The results show that in many cases, participants consistently chose the sentence with one AUX over the other. In other words, not all DAVs in our sample were perceived as such for our participants. If one establishes some arbitrary cutoff points along the continuum of variation in participants’ responses, one can divide DAVs into three even groups. A total of 12 DAVs out of 32 showed no or negligible (2%<) variation in participants’ choices, that is, they were only considered acceptable with one of the two AUX. These “non-DAVs” are aumentare, diminuire, funzionare, ingrassare, sopravvivere, sorgere, appartenere, durare, marcire, trionfare, continuare, and migliorare. It is worth noticing that most of these “non-DAVs” featured AUX E. Among the remaining 20 DAVs, 10 showed a low degree of variation (20%<), i.e., aderire, peggiorare, rotolare, tremare, correre, prevalere, atterrare, brillare, resistere, and tentennare, and the remaining 10 showed a high degree of variation (>20%), i.e., cedere, convivere, avanzare, circolare, germogliare, saltare, decollare, fiorire, inciampare, and sbandare.

Table 8 reports participants’ choices between competing stimuli (featuring either A or E) by pools.

Table 8:

Participants choices between competing stimuli by pools (percentage).

Pool Condition % featuring E % featuring A
Alpha atelic 71 29
telic 79 21
Beta Animate 45 55
Inanimate 40 60

At a glance, in pool alpha, the preferred stimuli were those containing AUX E, apparently regardless of verb telicity. On the contrary, in pool beta, the preferred stimuli were those containing AUX A, apparently regardless of subject animacy. Yet, the preference for AUX E in pool alpha was much bigger than the preference for AUX A in pool beta.

7.2 RQ1: Did the Presence of a ±Telic Completion Influence Participants’ Choice in the Case of Stimuli Featuring Change of State and Stative DAVs?

A binary logistic regression (BLR) (function glm, package “car”, software R version 4.1.0) was employed to predict participants’ choices between stimuli featuring AUX A and E, given our set of predictors (categorical or continuous variables).[19]

In order to answer RQ1, we computed the probability that the binary outcome (participants’ choices between competing stimuli) in pool alpha (containing state and change of state verbs) was significantly affected by the presence of a ±telic sentence completion.[20] The BLR showed that semantics alone, i.e., the telic versus atelic contrast, did not influence participants’ choices with respect to pool alpha DAVs (p = 0.9). As also shown in Table 8, the model confirmed that E was the preferred AUX for stative and change of state DAVs, irrespective of verb semantics. However, although semantics was not significant as a main factor for the choice of both A and E, its interaction with AUX was. Namely, the impact of semantics was selective and concerned AUX E, but not AUX A. This means that when E was selected—but not when A was selected—semantics did impact the choice. In the presence of telic completions, the likelihood of participants choosing stimuli containing AUX E increased significantly (df 1, z = 3.703, p = 0.0002) (upper part of Figure 3). By contrast, the same interaction was not found with atelic completions, whose presence did not significantly increase the probability of choosing AUX A (lower part of Figure 3).

Figure 3: 
Interaction between completion type (i.e., telic vs. atelic) and the choice of AUX A (blue) and AUX E (orange).

Figure 3:

Interaction between completion type (i.e., telic vs. atelic) and the choice of AUX A (blue) and AUX E (orange).

7.3 RQ2: Did the Presence of ±Animate Grammatical Subjects Affect Participants’ Choice in the Case of Stimuli Featuring Nonmotional Process DAVs?

The animacy of the grammatical subject in isolation (as the main factor) did not significantly influence participants’ choices in pool beta (p = 0.4). AUX A, in fact, was the most chosen AUX among DAV of pool beta (nonmotional processes), irrespective of the animacy of the subject. Also, in this case, however, there was a significant interaction between AUX and verb semantics, but, again, this interaction only concerned AUX E and not A. Indeed, in the presence of an inanimate grammatical subject, the probability of participants selecting a stimulus containing AUX E increased significantly (df 1, z = 2.074, p = 0.038) (lower part of Figure 4). The same interaction did not hold between AUX A and animate subjects (upper part of Figure 4).

Figure 4: 
Interaction between animacy of the grammatical subject and the choice of AUX A (blue) and AUX E (orange).

Figure 4:

Interaction between animacy of the grammatical subject and the choice of AUX A (blue) and AUX E (orange).

These asymmetries between the impact of semantics in E and A in both pools alpha and beta are discussed in Section 8.

7.4 RQ3: Did Raw Frequency of AUX + PP Combinations in the Input, Multiple Association Scores, and Unidirectional BTPs Between AUX and PP Influence Participants’ Choices Among Competing Stimuli?

Frequency of AUX + PP combinations in the input, association scores, and BTPs all straightforwardly affected our participants’ choices. The more frequent a given AUX + PP combination was in the input, the more strongly associated a given AUX and the PP were, the higher the BTP between the PP and a given AUX were, the more likely participants were to select a sentence stimulus containing that AUX (Frequency: df 1, z = 19.53, p = ≈0; LogDice: df 1, z = 24.27, p = ≈0; Tscore: df 1, z = 22.01, p = ≈0; BTP: df 1, z = 30.97, p = ≈0).

It should be recalled that in our verb sample, AUX E was more frequent than AUX A, in general. Moreover, AUX E was predominant with verbs of state and change of state (pool alpha), whereas AUX A was more frequent with verbs of nonmotional processes (pool beta) (Section 6.1). The BLR also confirmed that—although the affiliation to pool beta decreased the probability of choosing E over A by four points on average (df 1, z = −20.571, p ≈ 0)—this probability increased by 1.5 points on average for any increase of one (in a million) occurrence of AUX E + PP (df 1, z = 9.391, p ≈ 0). This suggests that the frequency of any AUX + PP combination affected participants’ choices independently of the frequency of AUX E and A, respectively, in “stative/change of state” or “nonmotional process” DAVs. The steeper lines in the upper right and lower left panels of Figure 5 visually represent the existence of a counterbalancing effect of the frequency of AUX + PP combinations in contexts, where AUX A is predominant (pool beta) and in contexts where AUX E is predominant (pool alpha).

Figure 5: 
Interaction between animacy of the grammatical subject and participants’ choices.

Figure 5:

Interaction between animacy of the grammatical subject and participants’ choices.

The effects of the association between AUX and PP sometimes differed depending on AUX type or on DAV semantics. Specifically, LogDice did not differ its impact depending on verb semantics, whereas Tscore and BTP positively affected participants’ choices only in pool alpha (telic vs. atelic contrast). The frequency of AUX + PP combinations in the input affected participants’ choices in all semantic conditions, except when the subject of the sentence was inanimate (bottom-left panel of Figure 6).

Figure 6: 
Interaction between frequency and semantic conditions.

Figure 6:

Interaction between frequency and semantic conditions.

7.5 RQ4: Did Resonance (i.e., the Presence of Transitive or Pronominal Counterparts of DAVs) Influence Participants’ Choices among Competing Stimuli?

The BLR showed that having a pronominal counterpart increased significantly the probability of choosing a sentence featuring AUX E (df 1, z = 16.49, p ≈ 0) while at the same time, decreasing the probability of selecting AUX A, though to a lesser extent (df 1, z = −2.28, p = 0.02). On the contrary, having a transitive counterpart did not increase the probability of selecting a sentence featuring AUX A nor did it decrease the probability of selecting AUX E (p = 0.409).

8 Discussion

There are four significant findings in this study. The first one is that in 40% of our stimuli featuring DAVs, participants consistently chose the sentence with one AUX over the other. In other terms, at least 40% of the sampled DAVs may be considered acceptable with both AUX only by linguists, but not by our participants, who accepted stimuli containing verbs diminuire, ingrassare, sopravvivere, appartenere, durare, marcire, and migliorare exclusively with AUX E and verbs trionfare and continuare exclusively with AUX A; other four verbs (aumentare, funzionare, sorgere, and atterrare) showed totally negligible oscillations between AUX choice. One may object that the participant sample was not representative of the native speaker population. A possible reply is that our participants’ choices largely reflected a tendency in larger population, as it can be seen by looking at large corpora of spoken and written contemporary Italian. Indeed, most selected AUX were 50–100 times more frequent in both CORIS and ItTenTen corpus than the unselected ones cf. (Table 3). This was not always the case though. For example, although cedere with AUX A is almost 400 times more frequent than with AUX E, our participants chose E 26% of times. Also, due to space constraints, an elaboration on the relationship between language use (and related corpora) and native speakers’ language competence is not possible. The data from this study can only suggest that among highly educated young adult native speakers living in Northern Italy, there could be less optionality in the choice of AUX than one that may be led to think when reading linguistics papers and manuals. Sorace (2000: 866, fn. 13) wrote that disagreement among native speakers is exactly what is expected for noncore verbs. Yet, in the present study, disagreement among our native participants concerned only the 60% of ASH noncore (peripheral) verbs. The remaining 40% would rather occupy the extremities of the ASH gradient (where A or E are selected without exceptions), virtually leaving a big hole in the middle.

The second result of the study is that the contribution of verb semantics to AUX selection in our data differed between AUX E and A. On the one hand, telic completions and the presence of inanimate subjects exerted their influence on participants’ choices, by increasing the likelihood that AUX E was selected even in contexts where A was prevalent (pool beta). On the other hand, the presence of atelic completions and the animacy of the subjects never increased the likelihood of AUX A being selected. These results may indicate that the choice of AUX E in both pools (stative/change of state and nonmotional process DAVs) was sensitive to semantic factors, whereas the choice of AUX A was not. Another explanation of the asymmetry between the impact of semantics on the choice of A and E may reside in the composition of pools themselves. Since the results showed that semantics impacted only on E in both pools, one could argue that the theoretical distinction between stative/change of state and nonmotional process DAVs is not confirmed by experimental data and has no correspondence in processing.

The third significant result of the study is that frequency and association scores strongly influenced the processing of DAVs across the board. The frequency-association bound largely overwrote both semantics and syntax because their effect was independent of both telicity and animacy, and of the distinction between states and processes. In general, AUX E was more sensitive to frequency and association scores than AUX A. Namely, Tscore and BTP positively affected participants’ choices only in pool alpha (telic vs. atelic contrast), where AUX E was prevalent. By contrast, the frequency of AUX + PP combinations in the input affected participants’ choices across all semantic conditions, except when the subject was inanimate. In general, the fact that AUX A was more impermeable especially to association scores is not easily interpretable. It might depend on the choice of verbs in either pools or on other factors (maybe also the competition between AUX A and the lexical uses of avere “have” as a verb expressing possession). This result could open interesting and promising directions for future research concerning the existence of an asymmetry underlying the processing of the auxiliation system in modern standard Italian.

Finally, our data suggest that having a pronominal counterpart (but not a transitive counterpart) was a significant factor for the choice. This means that—when selecting E—participants were probably considering whether or not the DAVs had a pronominal counterpart in Italian, and they probably elaborated on the similarities between the unaccusative and pronominal configurations at some abstract level. What I have dubbed “resonance effect” could therefore be considered a psychological realis, a processing factor whose presence and role at the syntax-semantics interface should perhaps be taken into account in future research on AUX selection.

Three shortcomings of this study should also be highlighted. The first is that participants were all university students, and because of this, our findings, perhaps, should not be generalized beyond young educated Italian native speakers who have been living in a Northern Italy city for many years. The second shortcoming is that participants may have become aware of the target feature during the trial and may have developed task-related strategies. However, an oral feedback survey conducted immediately after the trial revealed that such heuristics largely varied across both participants and verbs and were not adopted systematically, hence they may not have biased the results. The third shortcoming is that the timed forced-choice technique is an offline method and, as such, it cannot tap into implicit knowledge, but only into declarative knowledge, i.e., the memory of notions learned in school combined with cold reasoning and inferences. There is, however, (at least) one advantage in utilizing an offline task in order to tap into native speakers’ knowledge of AUX selection. Free from the pressure typically exerted by tasks measuring reaction times, participants in our study might have had more time (up to 9000 ms) to carefully consider not only the meaning of the verb, but also all the elements in its syntactic surrounding. These are precisely the factors that—given the theoretical background—were expected to codetermine participants’ responses with respect to AUX selection.

Corresponding author: Stefano Rastelli, University of Pavia, Laboratorio di Linguistica e Glottodidattica Sperimentale (LLEGS), Palazzo San Tommaso, Piazza del Lino 2, Pavia, 27100, Italy, E-mail:


Aranovich, Raul (eds.). 2007. Split auxiliary systems. A cross-linguistic perspective. Amsterdam-Philadelphia: John Benjamins.10.1075/tsl.69Search in Google Scholar

Alexiadou, Artemis, Elena Anagnostopoulou & Martin Everaert (eds.). 2004. The unaccusativity puzzle: Explorations of the syntax-lexicon interface. Oxford: Oxford University Press.10.1093/acprof:oso/9780199257652.001.0001Search in Google Scholar

Bard, Ellen Gurman, Dan Robertson & Antonella Sorace. 1996. Magnitude estimation of linguistic acceptability. Language 72. 32–68. in Google Scholar

Bard, Ellen Gurman, Cheryl Frenck-Mestre & Antonella Sorace. 2010. Processing auxiliary selection with Italian intransitive verbs. Linguistics 48(2). 325–362. in Google Scholar

Bentley, Delia. 2006. Split intransitivity in Italian. Berlin-New York: Mouton De Gruyter.10.1515/9783110896053Search in Google Scholar

Bentley, Delia & Thórhallur Eythórsson. 2003. Auxiliary selection and the semantics of unaccusativity. Lingua 114(4). 447–471. in Google Scholar

Borer, Hagit. 2004. The grammar machine. In Alexiadou-Artemis Anagnostopoulou & Elena-Everaert Martin (eds.), The unaccusativity puzzle. Exploration of the syntax-lexicon interface, 288–331. Oxford - New York: Oxford University Press.10.1093/acprof:oso/9780199257652.003.0012Search in Google Scholar

Burzio, Luigi. 1986. Italian syntax. A government-binding approach. Dordrecht: Springer.10.1007/978-94-009-4522-7Search in Google Scholar

Cennamo, Michela. 2008. The rise and development of analytic perfects in Italo-Romance. In Thórhallur Eythorsson (ed.), Grammatical change and linguistic theory, 115–142. Amsterdam-Philadelphia: John Benjamins.10.1075/la.113.05cenSearch in Google Scholar

D’Alessandro, Roberta. 2017. When you have too many features: Auxiliaries, agreement and clitics in Italian varieties. Glossa: A Journal of General Linguistics 2(1). 50. 1–36 in Google Scholar

Dowty, David. 1979. Word meaning and Montague grammar. Dordrecht: Reidel.10.1007/978-94-009-9473-7Search in Google Scholar

Dowty, David. 1991. Thematic proto-roles and argument selection. Language 67(3). 547–619. in Google Scholar

Foley, William & Robert Van Valin. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press.Search in Google Scholar

Folli, Raffaella & Heidi Harley. 2005. Flavors of v. In Paula Kempchinsky & Roumyana Slabakova (eds.), Aspectual inquiries. Studies in natural language and linguistic theory, Vol. 62. Dordrecht: Springer.10.1007/1-4020-3033-9_5Search in Google Scholar

Grassi, Corrado, Alberto Sobrero & Tullio Telmon. 1997. Fondamenti di dialettologia italiana. Roma-Bari: Laterza.Search in Google Scholar

Jakubíček, Milos, Adam Kilgarriff, Vojtech Kovář, Pavel Rychlý & Victor Suchomel. 2013. The TenTen corpus family. In 7th international corpus linguistics conference CL, 125–127.Search in Google Scholar

Jezek, Elisabetta. 2003. Classi di verbi tra semantica e sintassi. Pisa: ETS.Search in Google Scholar

Kailuweit, Rolf & Malte. Rosemeyer. 2015. Auxiliary selection revisited. Gradience and Gradualness. Berlin-Boston: De Gruyter.10.1515/9783110348866Search in Google Scholar

Kras, Tihana. 2010. Unaccusativity in L2 Italian at the Lexicon-syntax interface. In Michael Iverson (ed.), Proceedings of the 2009 mind/context divide workshop, 60–71. Somerville, MA: Cascadilla Proceedings Project.Search in Google Scholar

Lenci, Alessandro & Alessandra Zarcone. 2009. Un modello stocastico della classificazione azionale. In Linguistica e modelli tecnologici di ricerca: Atti del XL Congresso internazionale di studi della Società di linguistica italiana (SLI). Roma: Bulzoni. in Google Scholar

Levshina, Natalia. 2015. How to do linguistics with R. Amsterdam-Philadelphia: John Benjamins.10.1075/z.195Search in Google Scholar

Loporcaro, Michele. 2006. L’allineamento attivo-inattivo e il rapporto tra lessico e morfosintassi. In E. Cresti (ed.), Prospettive nello studio del lessico italiano. Proceeding of SILFI conference, Vol. I, 311–320. Firenze: FUP.Search in Google Scholar

Loporcaro, Michele. 2014. Perfective auxiliation in Italo-Romance. The complementarity of historical and modern cross-dialectal evidence. In Paola Benincà, Ledgeway Adam & Nigel Vincent (eds.), Diachrony and dialects: Grammatical change in the dialects of Italy. Oxford: Oxford University Press.10.1093/acprof:oso/9780198701781.003.0003Search in Google Scholar

Loporcaro, Michele. 2015. Perfective auxiliation with reflexives in medieval romance: Syntactic vs. semantic gradients. In Rolf Kailuweit & Malte Rosemeyer (eds.), Auxiliary selection revisited: Gradience and gradualness, 43–77. Berlin: De Gruyter.10.1515/9783110348866-003Search in Google Scholar

Maiden, Maiden & Cecilia Robustelli. 2000. A reference grammar of Modern Italian. London: Arnold Publisher.Search in Google Scholar

Malec, Wojciech. 2010. On the asymmetry of verb-noun collocations. In Janusz Arabsky & Adam Wojtaszek (eds.), Neurolingusitic and psycholinguistic perspectives on SLA, 126–143. Bristol-New York: Multilingual Matters.10.21832/9781847692429-011Search in Google Scholar

McClure, William. 1993. Unaccusativity and inner aspect. In Jonathan Mead (ed.), Proceedings of WCCFL 11. Stanford, Calif: CSLI.Search in Google Scholar

McFadden, Thomas. 2007. Auxiliary selextion. Language and Linguistics Compass 1/6. 674–708. in Google Scholar

Moens, Marc & Mark Steedman. 2005. Temporal ontology and temporal reference. In Inderieet Mani & Pusteiovskj James (eds.), The language of time: A reader, 93–114. Oxford, UK: Oxford University Press.Search in Google Scholar

Onnis, Luca, Heidi Waterfall & Shimon Edelman. 2008. Learn locally, act globally: Learning language with variation set cues. Cognition 109. 423–430. in Google Scholar

Perlmutter, David. 1978. Impersonal passives and the unaccusative hypothesis. Berkeley Linguistics Society 4. 157–189. in Google Scholar

Perlmutter, David. 1989. Multiattachment and the unaccusative hypothesis: The perfect auxiliary in Italian. Probus 1. 63–119. in Google Scholar

Pinkster, Harm. 1987. The strategy and chronology of the development of future and perfect tense auxiliaries in Latin. In Martin Harris & Paolo Ramat (eds.), Historical development of auxiliaries, 193–224. Berlin-New York: Mouton De Gruyter.10.1515/9783110856910.193Search in Google Scholar

Ramat, Paolo. 1987. Introduction. In Martin Harris & Paolo Ramat (eds.), Historical development of auxiliaries, 3–21. Berlin-New York: Mouton De Gruyter.10.1515/9783110891249.1Search in Google Scholar

Rohlfs, Gerhard. 1969. Grammatica Storica della lingua italiana e dei suoi dialetti. Sintassi e formazione delle parole. Torino, Einaudi.Search in Google Scholar

Rosen, Carol. 1987. Star means bad: A syntactic divertimento for Italianists. Italica 64/3. 443–476. in Google Scholar

Rosen, Carol. 1997. Auxiliation and serialization: On discerning the difference. In Alex Asina, Joan Bresnan & Peter Sells (eds.), Complex predicates, 175–202. Stanford, CA: CSLI.Search in Google Scholar

Sabatini, Francesco. 1985. L’italiano dell’uso medio”: Una realtà tra le varietà linguistiche italiane. In Gűnter Holtus & Edgar Radtke (eds.), Gesprochenes Italienisch in Geschichte und Gegenwart, 154–184. Narr: Tübingen.Search in Google Scholar

Salvi, Giampaolo & Laura Vanelli. 2004. Nuova grammatica Italiana. Bologna, Il Mulino.Search in Google Scholar

Schwarze, Christoph. 2009. Grammatica della lingua Italiana. Roma: Carocci.Search in Google Scholar

Serianni, Luca. 1991. Grammatica Italiana. Italiano comune e lingua letteraria. Torino: UTET.Search in Google Scholar

Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76(4). 859–890. in Google Scholar

Sorace, Antonella. 2004. Gradience at the lexicon-syntax interface: Evidence from auxiliary selection and implications for unaccusativity. In A. Alexiadou, E. Anagnostopoulou & M. Everaert (eds.), The unaccusativity puzzle. Exploration of the syntax-lexicon interface, 243–268. Oxford - New York: Oxford University Press.10.1093/acprof:oso/9780199257652.003.0010Search in Google Scholar

Sorace, Antonella. 2015. The cognitive complexity of auxiliary selection: From processing to grammaticality judgements. In Rolf Kailuweit & Malte Rosemeyer (eds.), Auxiliary selection revisited. Gradience and gradualness, 23–42. Berlin-Boston: De Gruyter.10.1515/9783110348866-002Search in Google Scholar

Thompson, Susan & Elissa Newport. 2007. Statistical learning of syntax: The role of transitional probability. Language Learning and Development 3(1). 1–42. in Google Scholar

Travis, deMena Lisa. 1994. Event Phrase and a theory of functional categories. In Paivi Koskinen (ed.), Proceedings of the 1994 Canadian linguistic association meeting at the University of Calgary, 559–570. Toronto: Working Papers in Linguistics.Search in Google Scholar

Travis, deMena Lisa. 2010. Inner aspect: The articulation of VP. Dordrecht: Springer.10.1007/978-90-481-8550-4Search in Google Scholar

van Hout, Angeliek. 2004. Unaccusativity as telicity checking. In Artemi Alexiadou, Anagnostopoulou Elena & Martin Everaert (eds.), The unaccusativity puzzle, 60–83. Oxford: Oxford University Press.10.1093/acprof:oso/9780199257652.003.0003Search in Google Scholar

van Valin, Robert. 1990. Semantic parameters of split intransitivity. Language 66(2). 221–260. in Google Scholar

Vernice, Mirta & Antonella Sorace. 2018. Animacy effects on the processing of intransitive verbs: An eye-tracking study. Language, Cognition and Neuroscience 33/7. 850–866. in Google Scholar

Published Online: 2022-08-03

© 2022 the author(s), published by De Gruyter, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.

Scroll Up Arrow