Crosslinguistic differences on the Present Perfect Puzzle: An experimental approach

. In this paper, we analyze how different temporal and referential properties of past-referring adverbials –specifically, hodiernality and deixis– are partially responsible for the crosslinguistic distribution of PAST and PERFECT markers across Dutch, Spanish, and English. To that end, we conducted an acceptability judgment task, where 160 subjects per language rated context-sentence pairs that display either a PAST or a PERFECT marker, and a temporal adverbial that is: (i) either temporally close to or temporally far from the speech time, and (ii), either deictic or not deictic. Results show that: (a) Dutch allows for its PERFECT marker to combine with any past-referring temporal adverbial, (b) Spanish only allows its PERFECT marker to combine with adverbials that locate the event temporally close to speech time, regardless of deixis, and (iii) that English prefers its PAST marker in all past-referring situations, but allows its PERFECT to combine with adverbials that are both deictic and temporally close to speech time, particularly when the adverb specifies an interval that is included in the day of utter-ance (e.g., this morning ), as opposed to adverbs that describe an interval that includes it (e.g., this month ).


Introduction.
The distribution of the Present Perfect 1 and the Simple Past in English is said to encompass a distinction between past events that have some current relevance or might continue into the present -which are expressed using the Present Perfect-, and past events that started and ended before utterance time -expressed with the Simple Past.This contrast is especially observable when the events at issue are anchored to a definite or specific time in the past.Since Klein (1992), it is well known that the Present Perfect marker cannot combine with a temporal adverbial referring to the past, so that the Simple Past has to be used instead, as in (1): (1) Chris *has left / left today at three o'clock.
(adapted from Klein 1992: 546, (ex.45)) In the sentence in (1), the specification of the past time reference ('today at three o'clock') makes the use of the Present Perfect form has left impossible, so that the only available option is the Simple Past form left.This incompatibility has been referred to in the literature as the Present Perfect Puzzle, since both the Present Perfect and the Simple Past are able to locate an event in the past, making it somewhat surprising that one of these markers -the Present Perfect-is ultimately incompatible with past-referring adverbials.
Most solutions for this pattern invoke a Reichenbachian framework (Reichenbach 1947), where an additional time point -the Reference Time-is considered.According to Reichenbach (1947), both the Simple Past and the Present Perfect situate the time of the event (E) before the time of speech (S), but the difference between these tense markers resides in the position of reference time (R).In the case of the Simple Past, the reference time (R) coincides with the event time (E), producing a E,R < S configuration, where < indicates precedence in time.Conversely, the Present Perfect makes the reference time (R) coincide with speech time (S), such that the configuration becomes E < R,S.Therefore, since temporal adverbials are argued to modify or target the reference time (R), the Present Perfect becomes incompatible with past-referring temporal adverbials, because these adverbs cannot target or modify a reference time that coincides with speech time.
Other Western European languages, however, such as Italian (Squartini and Bertinetto 2000), French (Vet 1980(Vet , 1992) ) or German (Musan 2002, Schaden 2009), do allow their corresponding PERFECT markers (i.e., Passato Prossimo, Passé Composé, Perfekt) to combine with past-referring adverbials, showing that the aforementioned constraint does not hold crosslinguistically.In a translation from (1), ( 2) illustrates this for Italian, while (3) does so for French, and ( 4) for German:  The Dutch PERFECT marker, the Voltooid Tegenwoordige Tijd (VTT, henceforth), is not affected by this constraint either, as (5a) shows.Rather, as (5b) exemplifies, the PERFECT is said to be preferred over the Dutch PAST, the Onvoltooid Verleden Tijd (OVT, henceforth), in such contexts ( de Swart 2007, van  The dialectal varieties of Peninsular Spanish spoken in Madrid and its surroundings appear to reflect an intermediate point in its availability to combine the Spanish PERFECT marker -the Pretérito Perfecto Compuesto-with past-referring temporal adverbials.These dialects allow the PERFECT to express a past event anchored to a specific time in the past, as long as the event has occurred in a time that is included in the day of utterance (Harris 1982, Schwenter 1994, González et al. 2019, Fuchs & González 2022) As (6a) indicates, Spanish allows its PERFECT to combine with temporal adverbials (hoy a las tres 'today at three o'clock') that create the relation E,R ⊆ day(S); that is, a relation in which the reference time (R) coincides with the event time (E), and both of these times are properly included within the day of the speech time (S).Moreover, in these cases, Spanish not only seems to allow the use of this marker, but also to prefer it over the Spanish PERFECTIVE PAST, the Pretérito Indefinido, as in (6b).Conversely, when the event time (E) is anchored to a past reference time (R) before the day of utterance (S), creating a relation E,R < day(S), as in (7), with the adverb ayer 'yesterday', only the PERFECTIVE PAST -the Pretérito Indefinido-seems to be allowed, as in (7b), while the Pretérito Perfecto Compuesto in (7a) is said to be ungrammatical.That is why the Spanish PERFECT has been defined as a hodiernal (i.e., relating to the present day) past marker, a temporal distinction that has been shown to be at play with specific dedicated markers in some other languages and language families (e.g., Comrie 1976, Dahl 1985).
There are other properties of past-referring temporal adverbials that seem to play a role in their compatibility with PERFECT markers.For example, some work in English has provided indications that deictic adverbials (i.e., adverbials whose reference is calculated with respect to the speaker's time/space center of reference) behave differently with respect to their (in)compatibility with the English Present Perfect (e.g., Hitzeman 1995).Different from (1), the Present Perfect seems to be able to combine with deictic past-time referring adverbials that include speech time (S), such as this afternoon, as in ( 8): (8) Chris has left / left York this afternoon.This sentence is acceptable with the Present Perfect even if this deictic adverb could technically be temporally locating the event of Chris' leaving at the exact same time that the adverbial at three o'clock, as in (1), a sentence that is considered ungrammatical with the Present Perfect.The main account for this pattern is that deictic adverbs target a relation in which reference time (R) is situated at speech time (S), and therefore do not create an incompatibility with the temporoaspectual configuration of the PERFECT.However, the role of deixis in the compatibility of the Present Perfect with past-referring temporal adverbials has not been systematically tested either in English or in Spanish.
Additionally, the example in ( 8) presents an adverbial that is both deictic and hodiernal, but it is also possible that these properties play each an independent role.Parallel corpus work (van der Klis et al. 2022) actually indicates that standard written Peninsular Spanish not only allows its PERFECT to combine with temporal adverbials that locate the event within the day of utterance, but also with deictic adverbials such as este mes 'this month', which locate the event within an interval that is calculated from speech time and includes the day of utterance, but can also temporally locate the event before that day (e.g., an event that occurred this month might have happened twelve days ago).This difference in the temporal extension of deictic adverbials (i.e., whether the time span they define includes or is included in the day of utterance) has not been systematically explored either.
In the following sections, we present an experimental study on the acceptability of different past-time referring adverbials with the PERFECT and the PAST markers of English, Spanish, and Dutch -with the expectation that the latter works as a control language, since we predict its PERFECT to have no restrictions to locate events in the past. 3Our main goal is to clarify the relationship between the temporal and referential properties expressed through adverbials and the use of PERFECT and PAST markers crosslinguistically.To this end, we conducted an acceptability judgment task on the use of PERFECT and PAST markers in these three languages to express differently temporally located past events.We consider a twofold distinction of past-referring temporal adverbials: temporal proximity and deixis.First, ( 6) and ( 7) indicate variation between adverbials related to the day of utterance and those that are not, showing the relevance of temporal proximity.Second, ( 1) and ( 8) drive a distinction between deictic and non-deictic adverbials.We describe the specifics of the experimental methodology in Section 2, we present the results in Section 3, and we discuss them and advance a general conclusion in Section 4.

Methods.
2.1.MATERIALS.We investigate UK English, Peninsular Spanish, and Netherlandic Dutch 4 use of PERFECT and PAST markers in combination with different temporal adverbials that we distinguish crossing two independent variables described directly below: • Temporal Proximity (T), which in +T cases refers to adverbials that are related to the day of utterance by being included in it (e.g., this morning), overlapping with it (e.g., today) or including it (e.g., this month), and conversely, in -T cases designates adverbials such as last month, which do not include or are included in the day of utterance.• Deixis (D).This variable refers in +D cases to adverbials whose temporal reference is deictic in nature.For example, to place an adverb such as yesterday on the timeline, we need information about the speaker's temporal location at reference time.Conversely, -D adverbials, such as in November, can be more easily placed on the timeline independently from the speaker's center of reference. 5  3 Interestingly, the Dutch VTT only presents no restrictions when it deals with events in dialogue or single sentences.To refer to states in the Past, or to express narrative discourse (i.e., sequences of events in the past), Dutch also needs to resort to its PAST marker, the OVT (e.g., Le Bruyn et al. 2019). 4We decided to test these European varieties since the parallel corpus work that identified some of the constraints that seem to be at work in the distributions of PERFECT and PAST markers in these languages (e.g., Le Bruyn et al. 2019, van der Klis et al. 2022) based its claims on originals and translations in these dialectal varieties. 5Of course we claim this 'independence' of -D adverbs only in relation to the +D cases, since an adverb such as in November also usually refers to either the previous or the following November with respect to speech time.
Each of these variables has two levels, so that the experimental design, including the grammatical marker as an independent variable, consists of 8 conditions (2x2x2).We created 8 different contexts, for a total of 64 experimental stimuli, presented in Appendix A (only in English for reasons of space).We organized these stimuli in a Latin Square design, so that each participant only saw one experimental condition per context.We additionally created 96 fillers of three different categories: (i) cases that convey either an experiential or a pluractional eventive meaning, with adverbs such as once or twice, which in the three languages are usually expressed with the PERFECT instead of the PAST (e.g., de Swart, forthcoming); (ii) cases with wh-questions and temporal sluices, which are assumed to be only acceptable with the PAST in the three languages (see Tellings & Fuchs, in prep., for reporting of these results); and (iii) unrelated fillers on the use of PRESENT and PROGRESSIVE markers to express futurate readings.
Each stimulus was displayed separately and was accompanied by an introductory context.All experimental sentences convey an achievement to control for lexical aspect (since achievements have no temporal extension, and can be placed directly on the timeline, unlike accomplishments or activities).An example item in English is shown below, with its introductory context in ( 9) and the different possible experimental conditions in

PAST -T, -D in November
Table 1: Experimental conditions, illustrated with an example, crossing three independent variables: grammatical marker, temporal proximity (T) and deixis (D), with two levels each.
2.2.PROCEDURE.Previous work on this topic has mostly relied on introspection (e.g., Klein 1992) or production data such as corpora (e.g., van der Klis et al. 2022) or force-choice tasks (e.g., Schwenter 1994), where results are binary (presence or absence of a specific marker).Moreover, corpus data of specific marker-adverb combinations can be relatively scarce.Since large-scale judgment data can provide a more nuanced perspective than corpus data (Kepser & Reis 2005, Francis 2022), not only showing the best form in a given context, but also indicating preferences between forms, we decided to run an online acceptability judgment task in which participants rated individually presented sentences in a 5-point Likert scale.They also had to answer yes-no comprehension questions, which followed 75% of the items, included to check that participants were paying attention to the experimental stimuli that they had to rate.Each participant saw a total of 8 experimental stimuli and 16 fillers.The allotted time to complete the task was of 15 minutes, with Since we are interested in the distribution of these markers and their compatibility with different kind of adverbials in each of these languages, we performed separate statistical analysis per language.We subjected the data to linear mixed-effect analysis, which were performed with random intercepts for subject and item, and fixed effects for the interaction of grammatical marker, temporal proximity and deixis. 6The study was approved by the Faculty Ethics Assessment Committee of the Faculty of Humanities (FEtC-H) of Utrecht University (reference number: 20-249-03). 7This was particularly surprising in the case of Spain, since previous reports indicate a wider use of the PAST in some of the Northern regions (e.g., Azpiazu 2013Azpiazu , 2015)).We consider that the lack of variability in our study is probably due to the nature of the task, which allows to rate both markers as (un)acceptable.However, we recognize that further exploration with a more controlled recruitment process could provide results where that dialectal variation within Peninsular Spanish is observed.
In Dutch, we find a main effect of marker (χ2(2) = 32.117;p < .001),favoring the PERFECT over the PAST across all conditions (β = 0.8031; p < .001). Figure 1 shows the means across conditions in this language.Spanish shows a significant interaction of Temporal Proximity*Marker (χ 2 (1) = 47.12;p < .001),with no effect of deixis.In the -T condition, there is a main effect of marker (χ 2 (1) = 57.07;p < .001),favoring the PAST over the PERFECT (β = 1.353; p < .001),but in the +T condition, there is no significant effect of marker (χ 2 (1) = 0.016; p = .90).A summary in terms of means per experimental condition for Spanish is shown in Figure 2. Finally, in English, there is a significant effect of Temporal Proximity*Deixis*Marker (χ 2 (2) = 6.373; p < .05),and a main effect of Marker, favoring the PAST over the PERFECT in all conditions (β= 1.0875; p < .001).The interaction effect arises because there is less of a categorical difference in the +T,+D condition, but a post-hoc test with Tukey correction still shows the effect of Marker (β= 0.394; p = .035). Figure 3 presents a bar graph with the means per condition in English.An interesting, more nuanced result is also revealed in English when subdividing +T,+D adverbials by whether the adverb includes the day of utterance or is included in it.In those cases, we find a significant marker effect in the first case(χ 2 (1) = 6.7711; p <.01) favoring the PAST over the PERFECT (β= 0.5931; p < .001),but the effect disappears in adverbs included in the day of utterance (χ 2 (1) = 0.5942; p = .4408;PERFECT mean = 4.25; PAST mean = 4.38), where both markers produce ratings not significantly different.A bar graph showing this contrast is presented in Figure 4.In summary, Dutch speakers prefer the VTT -its PERFECT marker-over the OVT -its PAST marker-across the board.Spanish speakers accept the PERFECT when the adverb is linked to the present, but there is no preference for the Pretérito Perfecto Compuesto in the +T condition: the Pretérito Indefinido receives similar ratings.English speakers prefer the Simple Past in all conditions but they accept the Present Perfect with deictic adverbials, especially when the adverb is included in the day of utterance.
4. Discussion and general conclusion.This study had the objective of experimentally testing the acceptability of PERFECT and PAST markers in combination with various past-time referring adverbials in English, Spanish, and Dutch, with the ultimate goal of providing a more thorough and crosslinguistically valid account of the Present Perfect Puzzle (Klein 1992).Based on previous claims in the literature, we assumed that the dimensions of variability in the temporal and referential properties of adverbials that could have an effect on the acceptability of these markers were whether the adverb indicates a time span that is included in the day of utterance (i.e., hodiernality) and whether the time span is calculated from the speaker's center of reference (i.e., deixis).
Firstly, the results from our acceptability judgment task confirm the patterns previously reported in the literature in these languages (e.g., van der Klis et al. 2022 for Dutch;Schwenter 1994 for Spanish; Klein 1992, Hitzeman 1995 for English).Dutch allows its PERFECT to combine with any kind of past-referring adverbial, Spanish only allows the PERFECT form to appear when adverbials are hodiernal, and English prefers its PAST marker to make reference to past events, but allows its PERFECT to appear when the adverbs that it combines with are both deictic, and indicate a time span included in the day of utterance.
Additionally, our experimental data refines these generalizations with respect to the distribution of PERFECT and PAST markers across languages.The behavioral results not only provide wider empirical support for previous claims, but also deepen our understanding about the effect that temporal and referential properties expressed by past-referring adverbials might have on the crosslinguistic distribution of these markers, revealing some patterns previously undescribed.
In the case of Spanish, we did not only show that the Pretérito Perfecto Compuesto is allowed to express events that adverbials situate within the day of utterance (e.g., this morning), but also that these events can be situated further away in time, as long as the reference time is kept at speech time so that the time span that the adverb indicates is calculated from the deictic center (e.g., this month).Moreover, and contra previous descriptions in the literature (e.g., Schwenter 1994, Azpiazu 2013), our experimental data also revealed that the Spanish PERFECT may be the preferred form to express these hodiernal events, but the Pretérito Indefinido -the Spanish PERFECTIVE PASTis also acceptable in these contexts.
The English experimental data also provides an additional insight.The English Present Perfect is allowed to combine with adverbials that are both deictic and that signal an interval which is temporally close to speech time.However, to properly account for the patterns in the data, we need to make a distinction between 'proper hodiernality' and 'extended hodiernality', since English only allows its PERFECT to combine with deictic adverbials that are 'properly' hodiernal -that is, adverbials that indicate a time span included in the day of utterance-, and rejects the use of the PERFECT when the deictic adverbial indicates an interval that includes the day of utterance.This distinction had not been previously reported, and might be relevant for understanding the distribution of PERFECT and PAST markers more widely.
All in all, we conclude that, aside from providing results that specify a finer grained picture of the Present Perfect Puzzle from a crosslinguistic perspective, this study functions as an

Table 1 :
(9) Peter and Theresa are planning to go to a concert next weekend.Peter offers to go get the tickets later today, but Theresa tells him:

Table 2 :
Mean acceptability scores by experimental condition in each language.