1.1 The perfect-as-state hypothesis: Theoretical background

Historically, auxiliary constructions that would become the modern perfect in both Germanic and Romance languages denote three kinds of meanings, all of which are stative: stative possession (3), result states as consequences of previous telic events (4), and states of experience (5–6; all from the 13^th century):

What is common to these different uses is the existence of a state whose nature changes across time; and these patterns are attested across a wide variety of languages, including English (e.g., Bybee et al. 1994; Dahl & Hedin 2000; Nicolle 2012). Importantly, unlike lexical states, the perfect states are intricately tied and layered on top of the event they are derived from, whether it is enfriar(se) ‘to get cold’, cortar ‘to cut’, oír ‘to hear’, or andar ‘to walk’.

Today’s tense system seems to reflect these historical roots. As illustrated in (1–2) above, it has been claimed that the basic semantic difference between the past and perfect tenses is that, while the past tense merely signals the anteriority of a given event, the perfect tense asserts not only that the event occurred, but also, that there is a state that follows from it. Therefore, while the past tense refers to an event, the perfect refers to a state connected to a past event.³

There are cases where this derived state is intuitive, as in I have broken my glasses, built on telic predicates like break: The derived state is an immediate result state of the glasses being affected in a breaking event. In combination with atelic predicates, that is, predicates that do not have an inherent result, the nature of the state invoked by the perfect tense becomes more subtle, as in Julius’ having been to London in (2). This led Parsons (1990) to propose the more inclusive notion of “R-state”: a state that results from having experienced the event, and that holds forever by virtue of the event having taken place.

The idea that the basic meaning of the perfect is to introduce a state is supported by its behavior in stativity tests (Dowty 1979; Katz 2003). The perfect patterns with state predicates in tests that were originally used to distinguish dynamic eventualities (e.g., to kiss) from states (e.g., to love) in English (Katz 2003); here, we apply them cross-linguistically to English and Spanish.

First, stative predicates, unlike eventive predicates, tend to be incompatible with progressive aspect (but see Granath & Wherrity 2014), both in English (7–8) and Spanish (9–10). The perfect in both languages not only patterns with states, but seems to be stricter than stative verbs in its incompatibility with the progressive (11–12):

Second, states are not accepted in wh-cleft constructions. This is true of English (13–14) and Spanish (15–16); and again, the perfect tense patterns with states (17–18):

Third, the default interpretation of states when embedded under must is an epistemic reading (19, 21), while the default interpretation of dynamic events is deontic (20, 22). According to our intuitions, to obtain an epistemic reading, they need contextual and prosodic support. Again, the perfect tense seems to follow states in this test (23–24):

A fourth test claims that certain kinds of states (26, 28), unlike events (25, 27), are incompatible with the imperative (but see Lakoff 1966). As in the progressive test (11–12), perfect tense seems to be even less compatible with the imperative than stative verbs (29–30):

Lastly, states do not move the narrative time (31), whereas events (32) do (Dry 1983; Marx & Wittenberg 2024):

In (31), the state of the sky being clear and that of Laura’s experience of fear coincide, and we don’t assume any change. In (32), however, while the sky is clear (no change), Laura’s situation changes from not being on a journey to being on a journey. The perfect, just like a state, does not move the narrative either:

Table 1 summarizes the main results from the stativity tests, showing that the perfect exhibits semantic, syntactic, and discourse-related properties that are typically associated with states. At the same time, subtle differences observed with respect to lexical states, such as the more systematic incompatibility with the progressive, are to be expected since the state introduced by the perfect is of a different nature: it contains the reference to a past event.

Based on both the diachronic development of the perfect tense and arguments from these stativity tests, it has been proposed that the perfect tense renders reference to past events into reference to states derived from past events: In (1), we talk about the travelling event that Julius experienced; but in (2), we talk about Julius’ acquired property of having been to London.⁴

The interpretive differences between (1) and (2) can be seen to reflect a difference in the semantic structure of these forms: Several different analyses assume that both forms refer to a previous, bounded event, but only the perfect introduces a reference situation (Stowell 1996; Giorgi & Pianesi 1997; Jardón 2021; Schmitt 2001; Ramchand 2018). Hence the perfect (36), unlike the past (35), features two distinct situations: an event description (s₀), and a derived, asserted state (s’):

The denotation for the perfect in (36) takes an event description (35) and adds a situation s’ in which someone holds a state derived from the occurrence of the initial event in s₀. Consequently, the perfect incorporates the structure of the past representation (35), but also adds the semantics of non-eventive properties held by an individual (e.g. the property of being altruistic in (37)).⁵

In view of the diachronic evidence as well as the grammatical behavior, the linguistic literature has put forth predictions about the event construal evoked by the perfect: The claim that (1) has a subtly different meaning from (2) is a claim about different mental representations of a situation. But what is the evidence for such a difference in event construals?

The recent psycholinguistic literature is rich in experimental evidence for different event construals, but mostly based on aspect within the same tense. Imperfective aspect (usually, progressive) seems to lead to an event construal that focuses on the ongoing event, while perfective aspect directs the comprehenders’ attention to the endpoint of an event. This results in different patterns of reference, priming of objects and locations, memory of events, and object representations (e.g., Ferretti et al. 2007; 2009; Madden & Zwaan 2003; Magliano & Schleich 2000; Misersky et al. 2021; Wampler & Wittenberg 2023; Zhou et al. 2014). In contrast to the effects of aspect on event representation, however, the question of how event construals differ depending on whether events are described in perfect tense or past tense, has not been addressed from a psycholinguistic perspective.

One reason for this is that the event construal corresponding to both past and perfect tense is derived from concluded events in the past. Therefore, the stative interpretation of the perfect tense necessarily also includes the event as it was ongoing (s₀ in 36); the difference in event construal between those two tense forms is thus arguably even subtler than in the case of aspectual differences: the stative interpretation is layered on top of an event interpretation. In order to test the perfect-as-state hypothesis as a hypothesis of mental representation, we therefore need a sensitive tool that distinguishes the construal of a completed event from a state, to measure the perfect’s consequences on event construal.

In this paper, we use boundedness, and in consequence, individuability, as our tool: One important difference between the two construals of Julius’ travels in (1) and (2) is that while the travels described in (1) find their end on the train back from London, the property described in (2) has no natural boundary: Julius will forever keep the property of having been to London. This property also does not fundamentally change if he goes back to London, because while we can count his trips as individual events, they all integrate into the state of having been to London: Tense affects our intuition about temporal boundedness.

1.2 Boundedness in verbs and nouns

Boundedness is not only a notion that applies to verbs. On the contrary, it has been a central concept in the discussion of nouns as well: For instance, atomic count nouns, such as apple, have referents with well-defined spatial boundaries, whereas the referents of non-atomic mass nouns, such as applesauce, are spatially unbounded. In consequence, atomic nouns can be counted and individuated (I have three apples, my apple is different from yours); but applesauce cannot – it needs the help of a classifier (?three applesauces, but three cups of applesauce).⁶ On the other hand, applesauce can be divided into arbitrary units and be added to indefinitely, and we can keep referring to it as applesauce. However, a piece of an apple cannot be called an apple anymore: It is an apple slice, an apple core, or an apple peel (Bale & Barner 2009; Rips & Hespos 2015, and many others).

The parallel between nouns and events has been an ongoing topic in linguistics, predominantly in the domain of lexical verb semantics (Bach 1986; Casati & Varzi 2008; Hale & Keyser 1993; Harley 2005; Jackendoff 1991; Krifka 1992; Quine 1960; Rothstein 2008; Verkuyl 1972). Usually, it has been used to explain why some events can be broken up and still count as the same event, and others cannot. For example, running can and does happen in fits and starts: If you go for a run tonight, you may stop to tie your laces, drink some water, and chat with friends you encounter along the way. Still, you could not claim that you have gone on more than one run today: Your run, interrupted or not, counts as one and the same event. Other events, in contrast, do not count as the same event when they are broken apart: A bedtime hug stops when mom unwraps her arms from the child; if she re-wraps her arms, it counts as a second hug (Anderson, Garrod, & Sanford 1983; Hoepelman & Rohrer 1980; Carlson 1981; Magliano & Zacks 2011; Kiparsky 1998; Zacks & Tversky 2001; Wellwood et al. 2016).

Since these properties of individuation and boundedness correlate strongly with properties of verbal aspect and telicity (although these two are separate notions; Declerck 1989; Vanden Wyngaerd 2001), one of the fault lines to draw parallels between nouns and verbs has been between punctual and durative verbs.

There are several linguistic tests to distinguish punctual (35) from durative (36) verbs (Vendler 1957; Comrie 1976; Smith 1991), for instance by combining predicates with a temporal for-phrase to detect punctual events that trigger an iterated reading: While John’s jumping in (38) is typically coerced into a repeated event, John’s dancing in (39) is most naturally interpreted as a single prolonged event, without an iterative reading (but see for discussion Bott 2010; Wittenberg & Levy 2017; and many others).

Based on this test, punctual verbs are inherently temporally bounded, while durative verbs are unbounded.⁷ Zooming out from dynamic events onto eventualities in general, we can draw another parallel, this time to the event-state distinction: As we have seen above, while events are bounded in time, states are unbounded.

The parallel to our example of Julius’ travel is evident (also see Wittenberg & Levy 2017, for a different parallel in the event domain): A trip to London can be individuated, and it cannot be arbitrarily segmented. For instance, Julius’ trip was not a trip to London if the train got stuck in the Eurostar tunnel and never reached its destination; and it counts as a new, second trip if Julius travels again next year.

The state of having been to London, in contrast, cannot be individuated anymore: Its representation (36) has parallels to the representation of a property being held by someone, like ‘altruistic’ (37), although it can easily be supplemented by a new trip which adds to the event representations which the state of having been to London is derived from (it is an open phenomenological question whether this state could be segmented somehow). In addition, this state of having been to London is unbounded: It does not end when Julius boards the train back to the continent; he will keep this property until the day he dies.

In consequence, mass nouns, durative verbs, and states share the lack of boundedness and possibility of being individuated. We have also seen above that the perfect tense has been argued to parallel states, according to linguistic tests and diachronic evidence, giving rise to the perfect-as-state hypothesis. Therefore, we argue that the parallel from nouns to verbs can be extended into the tense system (Table 2). We argue that the parallels provide an entry point into asking the question: How do event construals based on past tense and perfect tense differ?

1.3 The current studies: Measuring individuation to understand the perfect tense

The current studies investigate whether describing events using perfect, as opposed to past, affects event construal: The perfect-as-state hypothesis would predict that while past tense gives rise to the construal of a past event, perfect tense elicits the construal of a state that is derived from that past event, and layered around it. This is in contrast to a null hypothesis expected from the Extended Now analysis of the perfect, which would not predict stative properties in the perfect.

To our knowledge, the foundational prediction of the perfect-as-state analysis, shared by several different approaches, has never been tested beyond the linguistic evidence described above, even though this analysis predicts that the tense difference should map onto different representations in people’s minds. Therefore, in order to test the perfect-as-state hypothesis, we need to target mental construals in addition to linguistic judgments such as the syntactic and semantic tests above.

Our solution is to approximate this prediction by using event individuation as a tool. We argue that perfect tense, durative verbs, as well as mass nouns, all have in common that their referents cannot be as readily individuated as opposed to the referents of past tense, punctual verbs, and count nouns. With the perfect, we make an indirect assertion about the core eventuality⁸: The stative nature of this assertion predicts a more unbounded or undividuated interpretation of the event in question, regardless of predicate type.

This is in contrast to a null hypothesis expected from the Extended Now analysis of the perfect, which would not predict stative properties in the perfect. The unboundedness of the perfect according to the Extended Now analysis (see, e.g., Pancheva 2003), is computed from the combination of a predicate’s lexical type (telic/atelic) and a particular morphology (progressive/non progressive): to get an unbounded construal of a telic predicate like jump, one would have to add the progressive morphology: I have been jumping (vs. I have jumped).

In our studies, however, we control for morphology (we never show the perfect progressive). Furthermore, the Extended Now theory predicts a more unbounded interpretation of the perfect only in combination with punctual/telic predicates like jump, not durative/atelic ones like dance: The idea is that, while the perfect can form a derived durative from a punctual event, it would necessarily leave the temporal properties of durative predicates unaffected.

We build our experiments on a previous study on event individuation. Barner, Wagner, and Snedeker (2008) used a quantity judgment task on event nouns, in which they manipulated nominal syntax (count vs. mass) and event type (durative vs. punctual). They presented participants with a set of vignettes, all in past tense, describing two characters performing actions that were prototypically unbounded and durative, such as dancing, and prototypically bounded and punctual, such as jumping. Participants were asked “Who did more ___?”. This question was phrased using a light verb construction, either in mass syntax (e.g., Who did more dancing/jumping) or count syntax (e.g., Who did more dances/jumps). To answer this question, participants had two choices: One character did more of the action in number of times, and the other character did more of the action in a different, pre-tested dimension (e.g., jumping higher, dancing longer). The number-based choice therefore served as a measure of individuation. Barner et al. (2008) found that punctual events (e.g., jumping/jumps) were very likely to be individuated, regardless of the syntax on the event noun. In durative events, mass syntax (e.g., dancing) corresponded to lower individuation rates than count syntax (e.g., dances).

In a first step (Experiment 1; see Table 3 for an overview of the current studies in relation to Barner et al. 2008), we replicated Barner et al.’s (2008) study in English, and extended it in two ways: First, we crossed language families by including European Spanish. In order to make meaningful comparisons, we used a language in which both the past and the perfect tense match the English meaning, unlike, for instance, French, which has a past and a perfect tense, but they map to different meanings (Squartini & Bertinetto 2000; Stowell 2008). For this reason, we decided on Spanish, a language in which the past/perfect distinction is assumed to operate in a similar fashion as in English (Martínez Atienza 2008; RAE-ASALE 2009).

In line with Barner et al.’s (2008) results, we predicted an effect of nominal syntax: Punctual events (e.g., jumping/jumps/saltar/saltos) should tend to be individuated, irrespective of nominal syntax. Durative events should show lower individuation rates in mass syntax (e.g., dancing/bailar), compared to count syntax (e.g., dances/bailas).

In a second extension to Barner et al. (2008), to prepare the question whether perfect tense leads to state interpretations, we added tense as a manipulation in Experiment 1: Half of our participants judged vignettes in past tense, and half, in perfect tense. The predictions for this tense manipulation were modest, based both on Barner et al.’s (2008) results as well on other previous work on eventive mass/count distinctions (Wittenberg & Levy 2017): Nominal syntax is a powerful mechanism to introduce or suppress individuation, whereas, as discussed above, the difference in event construal between past tense and perfect tense should be quite subtle: after all, both past and perfect include a reference to concluded events in the past. The effect of tense is therefore predicted to be comparatively smaller than the more straightforward effect of nominal syntax on individuation.

With Experiment 1 serving as a conceptual replication, we then moved on to Experiment 2, in which we tested the effect of tense on individuation without the modulation through nominal syntax, in both English and Spanish. Instead of using light verb constructions, such as asking Who did more jumps?, we used full verb forms in past tense (e.g., Who jumped/danced more? ¿Quién saltó/bailó más?) and perfect tense (e.g., Who has jumped/danced more? ¿Quién ha saltado/bailado más?).

Our prediction for Experiment 2 was that if perfect tense leads to stative interpretations, then perfect should lead to less individuation than past tense, based on the properties of boundedness described above. With the number-based choice serving as a measure of individuation, the prediction from the Perfect-as-State analysis is that the perfect would lead to more unbounded readings of events, measured by participants’ choice of dimensions other than number: these include duration, but also distance, weight, length, or height.⁹

In addition, however, our study also addresses a question that Barner et al. (2008) could not: Namely, what the default individuation patterns are, depending on event type, without the factor of nominal syntax. Barner et al. (2008) interpreted their findings in line with the ‘Number Asymmetry Hypothesis’ (Barner & Snedeker 2006): According to this hypothesis, count syntax specifies individuation, while mass syntax is underspecified.¹⁰

However, an alternative interpretation of their results is imaginable: Namely, that event type (durative vs. punctual) leads to different baseline individuation rates. That is, either mass syntax reduces individuation for duratives; in this case, we should find high individuation rates in the absence of a nominal syntax manipulation. Alternatively, count syntax introduces individuation to duratives; in that case, in the absence of nominal syntax, we should find low individuation in durative events. This is in line with the logic of Table 2: Duratives are by default unbounded and unindividuated, like mass nouns, states, and the perfect tense.

2.1 Methods

2.2 Results

Overall, 95.5% of the control trials were judged correct in English and 95.8% in Spanish, indicating that the task was easily understood. Participants’ mean responses are shown in Figure 1: In both English and Spanish, we replicated Barner et al.’s (2008) results, specifically, the strong asymmetry between mass and count syntax.

Figure 1

Mean individuation responses for English (left) and Spanish (right). In both English and Spanish, we replicated Barner et al.’s (2008) findings. For visual clarity, we only illustrate the significant effects of pairwise comparisons (* = significant; • = marginally significant); for other results, please refer to the text. Error bars represent Standard Errors.

In English (Figure 1, left), when the quantity judgment questions used count syntax (e.g., jumps/dances), participants were more likely to judge quantity based on the number of times an action was performed than based on the other dimensions provided (M_COUNT = 0.95, SD = 0.23). In short, they were more likely to individuate the critical action events. In contrast, when the quantity judgment question contained mass syntax (e.g., jumping/dancing), participants individuated less, judging quantity based on the other dimensions more often than on number (M_MASS = 0.40, SD_MASS = 0.49). This asymmetry was statistically confirmed as a main effect of nominal syntax (Df = 1, χ² = 481.49, p < 0.001).

As Barner et al. (2008), we found a main effect of event type (Df=1, χ²=368.89, p < 0.001): Vignettes with punctual action events (e.g., jump, M_PUNCTUAL = 0.94, SD_PUNCTUAL = 0.24) led to significantly more individuation judgments than vignettes with durative action events (e.g., dance, M_DURATIVE = 0.46, SD_DURATIVE = 0.50).

Furthermore, the interaction between event type and nominal syntax was significant (Df = 1, χ² = 118.36, p < 0.001): In mass syntax conditions, quantity judgments were mostly based on number when combined with punctual events (M_{MASS-PUNCTUAL} = 0.88, SD_{MASS-PUNCTUAL} = 0.32), but not when combined with durative events (M_{MASS-DURATIVE} = 0.06, SD_{MASS-DURATIVE} = 0.23). For count syntax conditions, on the other hand, there was no difference between punctual and durative events (M_{COUNT-PUNCTUAL} = 0.97, SD_{COUNT-PUNCTUAL} = 0.17, M_{COUNT-DURATIVE} = 0.91, SD_{COUNT-DURATIVE} = 0.28).

Crucially, with regard to the effect of tense, differences were only numerical, with perfect tense leading to slightly lower individuation rates (M_PERFECT = 0.69, SD_PERFECT = 0.46 vs. M_PAST = 0.72, SD_PAST = 0.45). However, this was not statistically reliable: Participants judged quantity based on nominal syntax, or event type, or a combination of the two, regardless of whether the question was phrased in the simple past or perfect. Therefore, neither did the main effect of tense (Df = 1, χ² = 2.29, p = 0.13) reach significance, nor did tense interact with any of the other predictor variables (Dfs = 1, χ²s < 1.44, ps > 0.23).

This pattern of results did not change when testing the main effects of tense and nominal syntax separately within each verb type (no main effects of tense: Dfs = 1, χ²s < 2.49, ps > 0.11, interactions: Dfs = 1, χ²s < 1.53, ps > 0.22). Here, we also only found main effects of nominal syntax (Dfs = 1, χ²s > 42.85, ps < 0.001).

Finally, planned pairwise comparisons did not reveal significant differences between tenses within each combination of verb type and nominal syntax (βs > –0.03, ts > –0.97, ps > 0.33 [ps = 1]). There was only a marginal difference between simple past and perfect tense in the mass durative condition (β = –0.03, t = –1.88, p = 0.06 [p = 0.24]).

Barner et al.’s (2008) original overall pattern of results was replicated by the Spanish experiment as well (see Figure 1, right): Participants were more likely to individuate for count syntax (M_COUNT = 0.79, SD_COUNT = 0.41), than for mass syntax (M_MASS = 0.40, SD_MASS = 0.49; main effect of nominal syntax: Df = 1, χ² = 492.82, p < 0.001). Furthermore, individuation was driven by punctual events (M_PUNCTUAL = 0.74, SD_PUNCTUAL = 0.44), but less so by durative events (M_DURATIVE = 0.42, SD_DURATIVE = 0.49; main effect of event type: Df = 1, χ² = 251.55, p < 0.001). For durative events, participants were only more likely to judge quantity based on number when quantity judgment questions used count syntax (M_{COUNT-DURATIVE} = 0.76, SD_{COUNT-DURATIVE} = 0.43), rather than mass syntax (M_{MASS-DURATIVE} = 0.15, SD_{MASS-DURATIVE} = 0.35). For punctual events, this asymmetry was less pronounced (M_{COUNT-PUNCTUAL} = 0.81, SD_{COUNT-PUNCTUAL} = 0.39 vs. M_{MASS-PUNCTUAL} = 0.66, SD_{MASS-PUNCTUAL} = 0.47). This was also reflected by a significant interaction between the two predictor variables (Df = 1, χ² = 142.57, p < 0.001).

Contrary to the English experiment, tense had a significant effect on people’s quantity judgments (main effect of tense: Df = 1, χ² = 4.42, p = 0.04): Across event types and nominal syntax, participants individuated less in the perfect condition (M_PERFECT = 0.56, SD_PERFECT = 0.50 vs. M_PAST = 0.61, SD_PAST = 0.49). This was more pronounced in count syntax, as reflected by the marginally significant interaction between nominal syntax and tense (Df = 1, χ² = 3.42, p = 0.06).

Analyzing the data separately for each verb type, there were main effects of syntax (Dfs = 1, χ2s > 56.26, ps < 0.001). The split analysis furthermore revealed a main effect of tense which was significant for punctual events (Df = 1, χ² = 4.32, p = 0.04), but not for durative events (Df = 1, χ² = 0.92, p = 0.34): In the punctual conditions, perfect reduced people’s tendency to individuate, so that quantity judgments were more often based on dimensions other than number than in the past tense conditions. However, likely due to reduced power in the planned pairwise comparisons, this difference was neither significant for the count-punctual conditions (β = –0.04, t = –1.30, p = 0.19 [p = 0.76]), nor for the mass punctual conditions (β = –0.044, t = –1.25, p = 0.21 [p = 0.84]). The only significant difference between tenses was in the count-durative conditions (β = –0.08, t = –2.20, p = 0.03 [p = 0.12]) which yielded a significant interaction between nominal syntax and tense in the durative conditions (Df = 1, χ² = 4.79, p=0.03), not in the punctual conditions (Df = 1, χ² = 0.07, p = 0.79).

2.3 Discussion of Experiment 1

In Experiment 1, we successfully replicated Barner et al. (2008)’s results in two languages, finding that both English and Spanish speakers quantified events in count syntax more than in mass syntax. This was primarily driven by event type, as significant interactions of nominal syntax and event type showed: Punctual events (e.g., jumping/jumps/saltar/saltos) resulted in high individuation rates, regardless of nominal syntax, but in durative events, mass syntax (e.g., dancing/bailar) corresponded to lower individuation rates than count syntax (e.g., dances/bailes). These results extend the empirical basis for the Number Asymmetry Hypothesis further, into Spanish.

There was a significant effect of tense in Spanish, while the English trend was only numerical. However, they both pointed into the predicted direction: less individuation in the perfect tense, compared to the past tense. The feebleness of the effect in English is in line with previous works that highlight the strong influence of nominal syntax on event individuation, possibly drowning out an effect of tense. Recall the prediction from Section 1.3 that the difference in event construal between past and perfect should be quite subtle in comparison.

The findings in Spanish, however, seem to go in a different direction, questioning the role of nominal syntax in that language. In the absence of further investigation, this might have been due to divergent morphosyntactic factors between the two languages: for example, the use of infinitives in lieu of –ing nominals in the mass conditions, or the fronting of the nominal in all target questions in Spanish. We will come back to these language-specific differences and how they may have affected the results in the General Discussion.

With Experiment 1 providing the empirical baseline, we removed nominal syntax in Experiment 2. In Experiment 1, we did not preregister a statistical comparison between English and Spanish because, theoretically, there was no reason to predict one: Spanish and English have been analyzed as having similar meanings for the past/perfect distinction (Squartini & Bertinetto 2000; Martínez Atienza 2008; RAE-ASALE 2009). However, visual inspection of our data suggested that Spanish participants were less likely to individuate overall. Since Experiment 2 also eliminated the differences in morphosyntactic framing necessary in Experiment 1 (e.g. Who did more jumps? vs. Saltos, ¿quién dio más?, i.e., Jumps: who did more?), we preregistered a cross-linguistic comparison.

3.1 Methods

3.2 Statistical Analysis

The statistical analyses were preregistered on OSF (see section on Data Availability). As in Experiment 1, we first set up a logistic regression model in the R statistical environment (i.e., glmer from the stats package, R Core Team, 2014) and performed model comparisons to assess the significance of the full model. Event type (i.e., durative vs. punctual) and tense (i.e., past vs. perfect) were sum-contrast coded as predictor variables (i.e., levels: 1/–1) and item as random intercept, using a minimal intercept-only structure to ensure model convergence (i.e., no random slopes). We used people’s quantity judgments based on individuation as the dependent variable. Second, we conducted planned pairwise comparisons between tenses across event types.

In addition, we conducted a cross-linguistic analysis by including language (English vs. Spanish) as a contrast-coded predictor variable in the logistic regression model as described above. To test the cross-linguistic predictions, planned pairwise comparisons assessed only the effects of language in each tense and event type condition (Bonferroni-corrections based on this number of comparisons are bracketed).

3.3 Results

Similar to Experiment 1, accuracy in the control trials was at ceiling: 97.6% of the English and 97.9% of the Spanish control trials were judged correctly. Participants’ mean judgment responses in Experiment 2 are shown in Figure 2: For the English data (left), participants judged quantity more often based on number in combination with punctual events than with durative events. That is, across tense conditions, vignettes with punctual predicates yielded a higher amount of individuation (M_PUNCTUAL = 0.82, SD_PUNCTUAL = 0.38) than vignettes with durative predicates (M_DURATIVE = 0.15, SD_DURATIVE = 0.36), as reflected by a main effect of event type (Df = 1, χ² = 1139.80, p < 0.001).

Figure 2

Mean individuation responses for English (left) and Spanish (right). For visual clarity, we only illustrate the significant effects of pairwise comparisons (* = significant; • = marginally significant); for other results, please refer to the text. Error bars represent Standard Errors.

Furthermore, participants were less likely to judge quantity in terms of the number of times an action was performed in the perfect conditions (M_PERFECT = 0.45, SD_PERFECT = 0.50) than in the simple past conditions (M_PAST = 0.52, SD_PAST = 0.50), as reflected statistically in a main effect of tense (Df = 1, χ² = 10.97, p = 0.001). While the interaction between event type and tense was not significant (Df = 1, χ² = 1.47, p = 0.22), pairwise comparisons revealed that the effects of tense were stronger combined with durative events (SD_{PAST-DURATIVE} = 0.39 vs. SD_{PERFECT-DURATIVE} = 0.33; β = 0.03, t = 3.06, p = 0.002 [p = 0.004]), whereas the difference between simple past and perfect was not significant for punctual events (SD_{PAST-PUNCTUAL} = 0.37 vs. SD_{PERFECT-PUNCTUAL} = 0.40; β = 0.02, t = 1.52, p = 0.13 [p = 0.26]).

In the Spanish data (Figure 2, right), we similarly found a main effect of event type such that, across tenses, vignettes reporting punctual events were more individuated (M_PUNCTUAL = 0.54, SD_PUNCTUAL = 0.50) than vignettes reporting durative events (M_DURATIVE = 0.12, SD_DURATIVE = 0.33; Df = 1, χ² = 689.12, p < 0.001). In contrast to the English data, there was no main effect of tense (M_PERFECT = 0.31, SD_PERFECT = 0.46 vs. M_PAST = 0.39, SD_PAST = 0.49; Df = 1, χ² = 1.10, p = 0.30), but the interaction between the two predictor variables was significant (Df = 1, χ² = 11.93, p < 0.001): Only in the punctual condition did the perfect tense lead to fewer number-based judgments than in the past tense, as pairwise comparisons showed (SD_{PAST-PUNCTUAL} = 0.49 vs. SD_{PERFECT-PUNCTUAL} = 0.50; β = 0.05, t = 3.78, p < 0.001 [p < 0.002]). There was no significant difference between tenses in the durative conditions (SD_{PAST-DURATIVE} = 0.31 vs. SD_{PERFECT-DURATIVE} = 0.34; β = –0.01, t = –1.33, p = 0.18 [p = 0.36]).

In the cross-linguistic analysis, we not only found main effects of event type (Df = 1, χ² = 2017.20, p < 0.001), and tense (Df = 1, χ² = 8.97, p = 0.003), but also a main effect of language (Df = 1, χ² = 149.61, p < 0.001): That is, across conditions, Spanish participants were less likely to base their quantity judgments on number (M_SPANISH = 0.35, SD_SPANISH = 0.48) than English participants (M_ENGLISH = 0.49, SD_ENGLISH = 0.50).¹³

Furthermore, there was a significant interaction between event type and language (Df = 1, χ² = 65.08, p < 0.001) such that punctual events got less individuated in Spanish (M_{SPANISH-PUNCTUAL} = 0.54, SD_{SPANISH-PUNCTUAL} = 0.50) than in English (M_{ENGLISH-PUNCTUAL} = 0.82, SD_{ENGLISH-PUNCTUAL} = 0.38), but durative events were more alike across languages (M_{SPANISH-DURATIVE} = 0.12, SD_{SPANISH-DURATIVE} = 0.33 vs. M_{ENGLISH-DURATIVE} = 0.15, SD_{ENGLISH-DURATIVE} = 0.36).

Finally, these language differences did not only weigh differently across event type, but also between tenses within each event type, as reflected by a significant three-way-interaction (Df = 1, χ² = 9.90, p = 0.002): Planned pairwise comparisons between languages showed that, in Spanish, punctual events were individuated significantly less than in English, both in past (β = 0.13, t = 11.40, p < 0.001 [p < 0.004]) and perfect tense (β = 0.16, t = 13.12, p < 0.001 [p < 0.004]). For durative events, on the other hand, Spanish speakers individuated events less often than English speakers only in the past condition (β = 0.04, t = 3.75, p < 0.001 [p < 0.004]), not in the perfect condition (β = –0.003, t = –0.43, p = 0.67 [p = 1]).

3.4 Discussion of Experiment 2

Experiment 2 tested the effect of tense on individuation in both English and Spanish using simple verb forms in past and perfect tense. As in Experiment 1, a main effect of event type confirmed that durative events give rise to less individuation than punctual events. This result indicates that the default behavior of durative verbs is not to give rise to individuated interpretation. We will return to this in the General Discussion.

Crucially, we also found an effect of tense, as predicted by the perfect-as-state hypothesis, but not the Extended Now hypothesis: Perfect tense led to less individuation across the board: This result manifested in English by a main effect of tense, and in Spanish, by an interaction between tense and event type; the absence of a main effect in Spanish is conceivably due to a floor effect in the durative condition, since Spanish speakers generally chose less individuation than their English peers: Comparisons across languages revealed that in Spanish, punctual events were individuated significantly less than in English, both in past and perfect tense, while Spanish speakers individuated durative events less often than English speakers only in the past condition, not in the perfect condition.

We also found a general difference between English and Spanish: Notably, across conditions, Spanish participants were less likely to individuate events compared to English participants. This difference between languages was more pronounced in punctual events than durative events.