Spanish lower and upper bounded change of state verbs: focusing on transitive experiencer object verbs

2.1 Preliminaries

As introduced in Section 1, it has been proposed that both eventive TEO and TCoS verbs possess a causative component. Starting with psych verbs, TEO verbs are ambiguous between three readings: stative, eventive, and agentive (cf. Arad 1998: 3–4; Landau 2010: 6). In Spanish, the stative reading (see (3a)) arises when the experiencer is marked in the dative case (cf. Fábregas et al. 2017; Marín 2015). In this reading, a change in the mental state of the experiencer is not entailed, but the holding of a psychological state about some property of the nominative argument. Sentence (3a) yields an interpretation that there is something about the stimulus that brings about Camila to be in the state of being bothered. The nominative argument is considered the subject matter or target of the emotion (SM/T, cf. Pesetsky 1995), which is non-volitional (cf. Fábregas et al. 2017). The dative stative construction of TEO predicates shows an unmarked OVS linearization. In contrast, the eventive (see (3b)) and the agentive (see (3c)) readings occur typically when the experiencer is marked in the accusative case,^[2] the unmarked word order being SVO (cf. Cuervo 2010; Fábregas et al. 2017; Jiménez-Fernández and Rozwadowska 2017; Machicao y Priemer and Fritz-Huechante 2018). Causation is assumed to be involved in the two latter readings (Alexiadou and Iordăchioaia 2014; Arad 1998; Landau 2010; Van Voorst 1992; see Section 2.2).

In this study, we consider the eventive reading of TEO verbs and hence prevent the accessibility to psych features related to the stative reading of the items (cf. Arad 1998; Landau 2010; Marín 2015; Temme and Verhoeven 2017, among others). In the eventive reading (3b), the causative layer accommodates a causer (cf. Pesetsky 1995) which entails a change in the mental state of the experiencer (cf. Alexiadou and Anagnostopoulou 2020; Alexiadou and Iordăchioaia 2014; Fábregas et al. 2017; Fritz-Huechante et al. 2020; Marín 2015), i.e., there is a transition into the psychological state denoted by the verb. In their eventive structure, TEO verbs are able to form verbal passives (cf. Landau 2010; Pesetsky 1995). This also holds for the Spanish TEO verbs tested here, with some having a more marked passive structure (e.g., deleitar ‘delight’) than others (e.g., molestar ‘bother’), which is nevertheless possible according to native speakers’ acceptability judgements (cf. Fábregas and Marín 2015), as emphasized by the comparative constructions in (4a) and (4b), respectively.

The availability to accommodate a causer (cf. (5a)) or an agent (cf. (5b)) as a subject is known to be a property of TCoS predicates (cf. Levin 1993; Martin 2020; McKoon and Macfarland 2000; Schäfer 2008; Wright 2002). Syntactically TCoS verbs share many properties with canonical transitive verbs, e.g., they form regular passive structures (cf. (6a), (6b)). Parallel to TEO verbs, the eventive reading of TCoS predicates is taken into account in this study.

Interestingly, in the perfective, TCoS predicates as well as a subset of Spanish TEO predicates are able to produce readings in which the change of state in the affected entity (theme vs. experiencer) does not need to take place in the presence of an agentive subject (cf. TCoS (5b), TEO (3c)). This phenomenon is known as non-culmination or failed-attempt reading (cf. Bar-el 2005; Beavers and Lee 2020; Demirdache et al. 2017; Demirdache and Martin 2015; Martin and Schäfer 2017; Singh 1999; Tatevosov and Ivanov 2009; for TEO verbs see Fritz-Huechante et al. 2020). In contrast, non-culminating readings are not (easily) available in the presence of an inanimate causer (cf. TCoS (5a), TEO (3b)), i.e., a change of state is entailed (cf. Agent Control Hypothesis, Demirdache and Martin 2015). One explanation for such readings is that agents and causers (non-agentive subjects) are introduced differently in the syntactic structure: agents are introduced in Voice _ag and causers in Voice _c (cf. Martin 2020). In other words, causers (e.g., ‘the storm’ and ‘the wine’) introduce an eventuality which causes the event denoted by the VP, i.e., Camila / the tablecloth are brought into a particular state: the state of being bothered in (3b) and the state of being stained in (5a). In contrast, agents (e.g., Pablo) do not introduce a further event and a change of state in the affected entity does not necessarily need to obtain, i.e., it is possible that Camila / the tablecloth are not brought into the state of being bothered / the state of being stained in (3c) and (5b), respectively. Since an animate subject referent is associated with agenthood properties (cf. Dowty 1991; Van Valin and Wilkins 1996, among others) and hence has the potential to allow a reading where a change does not take place, an inanimate causer is used in examples and experimental material throughout this work in order to emphasize the eventive reading of both TEO and TCoS predicates and observe the degree at which the object changes its state.

2.2 Lexical aspect and event structure

In Section 1, we mentioned that TEO verbs in their eventive reading have predominantly been analyzed as accomplishment or TCoS verbs (Alexiadou and Iordăchioaia 2014; Landau 2010). This is the same for Spanish (Vanhoe 2004: 378). However, a different analysis has also been proposed for these verbs as inchoative state causatives, contrasting with the existing view (cf. Fritz-Huechante et al. 2020; based on Fábregas et al. 2012; Fábregas 2015; Fábregas and Marín 2015; Marín and McNally 2005; Marín and McNally 2011). In this section, we will briefly discuss the event structure and aspectual properties of TEO predicates, showing that both analyses apply to different subsets of verbs.

In the traditional view, eventive TEO predicates denote causing events that lead to a change in the mental state of the experiencer (Alexiadou and Anagnostopoulou 2020; Alexiadou and Iordăchioaia 2014; Arad 1998; contrasting from Landau 2010 that argues that the eventive reading of TEO does not involve a change of state). In that sense, TEO predicates are bi-eventive (cf. Pesetsky 1995), as other non-psych verbs (e.g., TCoS dry), and their subject’s referent is causal, i.e., the causer triggers a change in the mental state of the experiencer (cf. Alexiadou and Anagnostopoulou 2020, see also Section 2). Spanish eventive TEO predicates also show a bi-eventive structure (cf. Fábregas and Marín 2015; Fritz-Huechante et al. 2020). A common test used to determine the bi-eventivity of complex predicates is the use of the adverb again. Again is ambiguous between two readings: (a) a repetitive reading, which refers to the repetition of the causing event, and (b) a restitutive reading, which makes reference to the “resumption of a previous state” (cf. von Stechow 1996; see Alexiadou and Iordăchioaia 2014 for the application of this test on Greek and Romanian EO verbs and similar results). In Spanish, the adverb otra vez ‘one more time’ gives rise to both interpretations. In the psych domain (cf. (7)), the repetitive reading of otra vez ‘one more time’ relates to the causing event that there is another time when the melody does something that would delight or calm Camila down. The adverb targeting the causing event can be highlighted by the continuation of the but-clause which yields an interpretation that a delighting / calming action was one more time performed by the causer which was not realized by Camila. Likewise, otra vez is able to target the experiential state, yielding an interpretation that Camila is in the state of being delighted or calmed down again after having been in a state of e.g., unhappiness or agitation, i.e., a restitutive reading.

Similarly, applying this test to TCoS verbs such as arrugar ‘wrinkle’ and secar ‘dry’ produces the aforementioned readings, parallel to TEO predicates. In the case of (8), otra vez ‘one more time’ conveys: (a) a reading that the causing event took place one more time, i.e., another time when the machine performs an action that would cause the clothes to be wrinkled or dry, and (b) a reading where the caused state is targeted, i.e., the clothes being wrinkled or dry again after having been in a different state (e.g., smooth/wet).

The otra vez test shows that the causative TEO predicates under investigation are consistent with causative TCoS predicates in their event structure, namely that both types are bi-eventive.

2.3 Telicity

Research on TCoS predicates has shown variable telicity of these predicates (cf. Kennedy 2011; Kennedy and Levin 2008; Rappaport Hovav 2014; Winter 2006). Likewise, telicity has proved to be a controversial feature of TEO predicates (cf. Alexiadou and Iordăchioaia 2014; Fábregas and Marín 2015; Tenny 1994, among others). In the present section, we will discuss our basic assumptions about telic and atelic readings of TEO predicates, which are similar to those of TCoS. As will be seen in Section 3, the distinct telicity readings can be associated with the different scale structures of the verbs in this study (cf. Kennedy and Levin 2008).

The eventive reading of causative TEO predicates involves an activity or process which is realized by a causer putting the experiencer in a particular mental state, and like accomplishments, these predicates lexicalize a result state (cf. Alexiadou and Anagnostopoulou 2020; Alexiadou and Iordăchioaia 2014). As other non-TEO verbs, the change of state can be of a gradual nature (e.g., dry), or it can constitute a non-gradual change, i.e., from not being in a state to being in a state, or from one value to another (e.g., wrinkle). Crucially, in the case of verbs of gradual change, the event denoted by the verb reaches its natural endpoint or telos once the result state obtains. Hence, these verbs are predominantly telic. In Spanish, serenar ‘calm down’ and despreocupar ‘make free of worries’ are instantiations of telic TEO verbs of gradual change. A common telicity test is the use of the progressive which licenses the inference to the perfect with atelic predicates, whereas the same is not the case with telic predicates (cf. Dowty 1991; Rothstein 2004, among others). For instance, an accomplishment predicate such as build a house in the progressive (9a) does not generate the inference that the house is already built (cf. (9b)). That is, the event has yet not reached its maximum telos which coincides with the house being completely built. The same is the case with the TEO predicate serenar ‘calm down’. Sentence (10a) shows that the ongoing event of calming down performed by the melody does not yield the interpretation that the experiencer is already in the state of being calm (10b). Native speakers agree that Camila is in the process of calming down, however, this process has yet not achieved its telos (see also Section 3).

Spanish TCoS predicates present the same behavior with respect to the progressive. The predicate secar ‘dry’ in (11a) shows that the event of drying is ongoing. A result state where the clothes are (completely) dry has not been achieved, hence an inference to the perfect is not possible (11b).

Verbs such as TEO deleitar ‘delight’ and TCoS arrugar ‘wrinkle’ constitute verbs of non-gradual change (see also Section 4). The change in the affected entity successfully starts at the minimal evidence of a result state with no obvious maximal state. Thus, in contrast to (10) and (11), these predicates are predominately atelic.^[3] For instance, the progressive with the verb arrugar ‘wrinkle’ in (12a) describes progressively larger subevents of wrinkling, licensing the inference to the perfect, i.e., the paper is already wrinkled (cf. (12b)). The same is the case for TEO verbs such as deleitar ‘delight’. Recent studies have proposed that a large group of TEO predicates makes reference to the onset of the experiential state they are associated with (cf. Fábregas 2015; Fábregas and Marín 2015; Fritz-Huechante et al. 2020, based on Marín and McNally 2011; Piñón 1997). Crucially, these predicates do not make allusion to the change that produces said state, but to a causative event that signals the starting (represented by a boundary) of the state, also known as inchoative state causative predicates. Applying the progressive test to these predicates, native speakers constantly interpreted sentence (13a) as an ongoing event where the state of being delighted is progressively enlarging with no maximum state (see also Section 3).

As seen in the previous sections, both TEO and TCoS predicates share some basic semantic characteristics in that they are bi-eventive consisting of a causing eventuality and a caused state. The causative layer is able to accommodate a causer argument which elicits the change or starting of the state. Two subtypes of verbs per verb class are also distinguishable: (a) those verbs that make reference to a process that leads to a culminating endpoint or telos, e.g., TCoS secar ‘dry’, TEO serenar ‘calm down’ (also defined as upper-bounded verbs, see Sections 3 and 4), and (b) those verbs that denote a change from a zero to a non-zero value (or the onset of a state) or from one value to another, e.g., TCoS arrugar ‘wrinkle’, TEO deleitar ‘delight’, respectively (determined as lower-bounded verbs, cf. Sections 3 and 4). These two groups of verbs also present differences regarding telicity readings, with upper-bounded verbs being typically telic whereas lower-bounded verbs are commonly atelic. In what follows, we will explain that an analysis in terms of scalar structure better explains the fine-grained distinction of TEO and TCoS verbs between upper- and lower-bounded lexical items. The meaning of the semantic core of these verbs as well as the identification of a standard degree of comparison (cf. Section 3) contribute to the distinction of the two different scales (cf. Section 4) for both psych and non-psych change of state verbs.

3.1 Gradability at the semantic core

The semantic core of TEO and TCoS verbs shows a stative attribute. This attribute makes reference to a gradable property which is represented by either a basic or a deverbal adjective (cf. Kennedy and McNally 1999, 2005; Rappaport Hovav 2014). The conventional meaning of the semantic core is one of the factors that accounts for the verb’s gradability. This core sets a range of ordered values which is the foundation to build a scale (cf. Hay et al. 1999; Kearns 2007; Kennedy and McNally 2005; Rappaport Hovav 2008; Rappaport Hovav and Levin 1998; Rotstein and Winter 2004). As previously mentioned, Spanish deleitar ‘delight’ and secar ‘dry’ are examples of verbs whose semantic cores describe a stative scalar attribute: the state of being delighted and the state of being dry, respectively. Degree modifiers, e.g., much, quite a lot, a little, etc., are commonly used to test whether a predicate is gradable. Such modifiers access the gradable property of a particular predicate and raise the degree the affected argument has at the beginning of the event (whatever this degree is at the beginning of such event) with respect to its property (Kennedy and McNally 2005; Piñón 2005). In (14), the modifiers un poco ‘a little’ and bastante ‘quite a lot’ add new degrees increasing the existing degrees of the property (the property of delight and the property of dryness) that the object (Camila and the clothes) holds at the beginning of the event. In sentence (14b), this new degree is not the necessary maximum degree that manifests the attainment of the property of dryness (see also Section 4), but what the modifiers specify. In the case of sentence (14a), since a change in the mental state of the experiencer has occurred which manifests the property of the semantic core of the verb, the new degree is an increase in the intensity of the emotion specified by un poco ‘a little’ and bastante ‘quite a lot’.

As seen so far, both TEO and TCoS predicates show gradable properties which are specified by their stative semantic cores. Important to our study is whether a standard degree of comparison is possible to be identified in the semantic core of the verbs which will set some parameters to form a scale in the verbal domain. This is the issue of the following section.

3.2 Standard degree of comparison

The standard degree of comparison is the degree that shows the attainment of the property stated at the semantic core or base of the verb (cf. Kennedy 1999; Kennedy and Levin 2008; Kennedy and McNally 2005). Scalar analyses argue that the attributes of the semantic core have an impact on the identification of the standard degree, i.e., whether the conventional meaning of the core is sufficient to establish this degree (also known as an absolute base) or whether the core meaning needs of contextual factors (e.g., domain of the discourse, interests and expectations of the discourse participants, etc.) to identify it (i.e., a relative base, cf. Kennedy and McNally 2005).^[5]

Kennedy (1999) and Kennedy and Levin (2008) define the standard degree as the degree that stands out relative to a property measured by the semantic core of a verb in some context. Once identified, the standard degree is subsequently introduced into the measure of change function (a function in the verbal domain, cf. Kennedy and Levin 2008, see Section 4) by means of degree morphology, i.e., a function represented by pos (a positive morphologically unmarked form), which applies to the semantic core of the verbs. In simple words, the standard degree function takes the base meaning of a verb to degrees and returns a property of events (for a detailed formal representation of pos and the standard of comparison, see Kennedy and Levin 2008; Kennedy and McNally 2005; based on Kennedy 1999; von Stechow 1984). This section deals with the identification of a standard degree of comparison in both verb classes considered in this study, TCoS verbs and TEO verbs, and further classifies them according to this feature.

4.1 Lower and upper bounds

Verbs such as TEO serenar ‘calm down’ and TCoS secar ‘dry’ show an upper-bounded scale structure. As mentioned previously, the measure of change function takes an object x and an event e and returns the amount that x changes (regarding the property specified at the verb’s semantic core) by participating in the event e in the form of a degree (cf. Kennedy and Levin 2008: 17). In the case of secar ‘dry’ in (24), the measure of change function takes the object ‘the clothes’ and measures its change in the event of drying, returning as an output a positive degree which captures the difference between the degree of dryness of ‘the clothes’ at the beginning of the event and the degree of dryness of ‘the clothes’ at the end of such event. Crucially, since a maximum standard degree is first identified in the semantic core of the verb (cf. Section 3.2.1), a maximal change on the scale (associated with the semantic core) is expected to take place in the event (unless otherwise specified, cf. (14b)), providing an upper bound for the scale. In sentence (24), the requisite for the clothes to be dry is that the degree of change in the clothes reaches the upper bound of the scale. Importantly, due to the definition of the measure of change function, we conceive the scale of these verbs as having both a minimal and a maximal bound. This is because the function establishes the obligatory presence of a minimum (and subsequent positive) value of change which maps onto the lower bound of the scale (cf. Kennedy and Levin 2008). The presence of an upper bound can be tested by the maximality modifier completely which has a maximal endpoint-oriented use and occurs felicitously with upper-bounded predicates (cf. Hay et al. 1999; Kennedy and McNally 1999, 2005). In sentence (24), the adverb completamente ‘completely’ indicates that the upper bound is reached and a continuation where higher degrees in the scale are reached is not possible. Native speakers constantly rejected the but-clause as the continuation of the sentence since this generates a contradiction.

Likewise, TEO verbs such serenar ‘calm down’ in (25) also show an upper-bounded scale structure. The measure of change function takes the experiencer and measures its change in the event of calming down. The function retrieves the positive difference degree of the measured property of calmness of the experiencer at the initiation and end of the event. As in secar ‘dry’, a maximal change on the scale is expected to take place (i.e., reaching the upper bound). This change is associated with the specifications of the semantic core, which in this case it relates to the total absence of the previous state (cf. Section 3.2.2). As with upper-bounded TCoS verbs, we conceive the scale structure of upper-bounded TEO verbs possessing both a minimal and a maximal bound. Applying the completely-test to these predicates (cf. (25)), we obtain similar intuitions as in (24) from native speakers. The maximality modifier entails that the end of the scale of calmness has been reached and subsequent higher degrees are not possible to reach, as shown by the continuation of the but-clause which generates a contradiction.^[9]

Another subset of verbs exhibits a lower-bounded scale structure which is open in its upper end. This is the case of TEO deleitar ‘delight’ and TCoS arrugar ‘wrinkle’. Like upper-bounded verbs, the measure of change function takes the object ‘the paper’ in (26) and the experiencer ‘Camila’ in (27) and gives back the positive difference degree of the measured property of the object at the beginning and end of the event, i.e., the event of delighting and the event of wrinkling, respectively. The presence of a lower bound in the scale determines that the object needs to reach the threshold (the standard degree) of the corresponding property to at least a minimum degree to manifest the attainment of such property (cf. Section 3.2).^[10] In contrast to upper-bounded verbs, no maximum degree needs to be reached for the predicate to hold true, hence no upper bound is identified in the scale. Due to the lack of an upper bound, the co-occurrence of the maximality modifier completely with these predicates is not felicitous (see (26) and (27)). Acceptability judgements are restored when the modifier gets reinterpreted as a synonym of very (cf. Hay et al. 1999; Kennedy and McNally 1999, 2005). Native speakers’ judgements coincided in that completamente ‘completely’ in both sentences (26) and (27) is used with the meaning of ‘very much’. The reinterpretation of the adverb makes reaching higher degrees in the scale possible and the continuation of the but-clause does not generate a contradiction.

Interestingly, the addition of extra contextual information, e.g., the adverb completely, might interfere with the natural lower-bounded reading of the verbs, i.e., an upper-bounded reading might also be available. Since completely is upper-bounded oriented, interpretations are formed by having in mind, e.g.,: (a) the size of the paper, in the case of arrugar ‘wrinkle’, and (b) the experiencer as a container or incremental theme, in the case of deleitar ‘delight’ (for an analysis of the experiencer as an incremental theme, see Willim 2020). For instance, in (26) a reference in mind could be set to a commonly used size of a sheet of paper (e.g., A4, see footnote 3, cf. Hay et al. 1999). Consequently, another possible interpretation of (26) is: the printer wrinkled the sheet of paper to its totality so that no more wrinkles can be added. In other words, the size of the paper aids to the identification of a possible upper bound in the scale. Similarly, in (27), a reference value associated with an upper bound could be Camila being delighted to the point of smiling. A possible interpretation of (27) would follow as: the story delighted Camila to its totality (i.e., reaching the point of smiling) so no more degrees need to be reached. As stated in Section 3.2.1 for upper-bounded items, in this study we avoid presenting the target verbs in rich context environments (see Section 5.2) in order to concentrate on the regular scalar readings of the verbs.

This section showed that the classification of TEO verbs in terms of their scale structure is pertinent, and crucially, it patterns with the existing scalar classification of TCoS verbs. Both upper-bounded TEO and TCoS verbs possess a scale where a maximum degree of change needs to be reached in order for the predicate to hold true. In contrast, lower-bounded verbs show a scale structure where the manifestation of only a minimum degree related to their corresponding property is sufficient to obtain a true predicate. Table 1 summarizes the scalar and semantic properties of TEO predicates presented so far in parallel to the features of TCoS predicates.^[11]

4.2 Degree modifiers and scale structures

The bounds of the verbs’ scale structure can be distinguished by means of degree modifiers (cf. Section 3.2). However, as mentioned earlier, the orientation of the degree adverbials might affect the likelihood that a speaker would yield an inference different from the natural interpretation of the scale of the item at hand.

Research on adjectives has shown that maximizer modifiers, e.g., completely, totally, are frequently used with upper-bounded items (cf. Hay et al. 1999; Kennedy 2007; Kennedy and McNally 1999, 2005, among others). For instance, x is completely A is true if x is A to the maximum degree related to a specific property. In the case of minimizer modifiers, e.g., slightly, a little, x is slightly A is true if x is A to at least some non-zero degree on the scale of A, hence frequently co-occurring with lower-bounded items. However, maximizer modifiers are not restricted to only upper-bounded items as well as minimizer modifiers do not only co-occur with lower-bounded items. A difficulty arises when these modifiers mismatch their natural co-occurrences. In the case of maximizers and lower-bounded items, these modifiers can force a maximal change reading of the sentence as seen in examples (26) and (27). Speakers can appeal to other factors, such as the size of the affected entity, as an attempt to repair the sentence. Likewise, in the case of minimizers and upper-bounded items, minimizers can yield an interpretation where the adverb (e.g., slightly) targets the upper bound of the verb instead of the lower-bound. For instance, sentence (28) can yield the natural lower-bounded interpretation where the clothes are more wet than dry, i.e., the clothes are a little dry, and a more restricted upper-bounded interpretation where slightly dry implies rather dry, i.e., the clothes are more dry than wet (for analyses on degree modifiers and scalar interpretations see McNally 2011; Sassoon and Zevakhina 2012, among others).

In this study, we focus on degree modifiers that do not force a maximal or minimal referential reading on the target items, such as almost. Research on adjectives has shown that almost is sensitive to the standard degree of comparison an item possesses (cf. Amaral and Del Prete 2010; Hitzeman 1992; Penka 2006; Rotstein and Winter 2004, among others). The adverb requires an adjective that has an identifiable standard degree in order to combine with it. The scope of almost is on the minimum or maximum degree of the pertinent scale of the adjective. It gives as an output an interval that is close by but lower than the standard degree, i.e., x is almost A entails x is not A (cf. Rotstein and Winter 2004: 267). This is also the case of the Spanish adverb casi ‘almost’. In co-occurrence with a maximum standard degree adjective (see (29)), the adverb yields a reading that Camila / the clothes are not calm / not dry but nearly calm / dry, i.e., casi ‘almost’ targets the degrees close and lower than the maximum standard degree.

In the case of minimum standard degree items, casi ‘almost’ yields an interpretation that x is not A at all (cf. Rotstein and Winter 2004: 267). In example (30), the adverb targets the values below the minimum standard degree, i.e., sentence (30a) yields an interpretation that Camila does not have the property of being delighted. Likewise, (30b) is interpreted as the paper does not have the property of being wrinkled. The co-occurrence of casi ‘almost’ with the adjectives in (30) shows to be a little odd since the adverb has scope over degrees that are not within the pertinent scale.

The following section deals with the experimental design that targets the different scalar readings the verbs in this study are able to produce considering the aforementioned features of the verbs.

5.1 Study design

In order to test the scalar structures of the predicates, we made use of the scalar adverbial casi ‘almost’, introduced in Section 4.2. Casi ‘almost’ is able to target the standard degree in the semantic core of the verbs. In the verbal domain, we observe that the adverb is able to target the bounds of the verbs’ scales, requiring the scale to be at least partially closed. Similarly to adjectives, with upper-bounded predicates as in (31), casi ‘almost’ selects values that are close and adjacent to the upper bound of the relevant scale. Sentence (31) is interpreted as ‘the circle was printed to a degree near completeness by the printer’.

In the case of lower-bounded predicates, the adverb has scope over the interval below the lower bound. Since the minimum degree of change (lower bound) refers to the minimal state that makes the predicate true, casi is interpreted such that the object does not reach the necessary minimum degree that manifests the property specified at the semantic core of the verb, hence there is no change in the object. This is exemplified by quemar ‘burn’ in (32). Casi ‘almost’ targets the interval below and adjacent to the starting of the state of being burned, thus the inference that obtains is that the wood was not burned by the fire.

Taking into consideration the semantic features of casi ‘almost’, we constructed sentences that contained verbs whose semantic core identifies a minimum standard degree, i.e., forming a partially closed scale, and verbs whose semantic core possesses a maximum standard degree, i.e., forming a totally closed scale. The classification of the verbs into these two sub-classes (lower-bounded vs. upper-bounded) was done by applying degree modifier tests such as the co-occurrence of the adverbs medio ‘half’, parcialmente ‘partially (cf. Section 3.2), and completamente ‘completely’ (cf. Section 4.1). For the upper-bounded set, we assumed that the upper bound is the prominent bound in the scale since it constitutes the value that manifests the attainment of the property specified by the semantic core of the verb and makes the predicate true. Furthermore, lexically TEO verbs are differentiated between a set of verbs which makes reference to the onset of a state (lower-bounded TEO verbs) and a set of verbs that denotes a culminating endpoint (upper-bounded TEO verbs), as outlined in Section 2.2. Similarly, TCoS verbs are classified as lower- and uppper-bounded items. The experiment design examines the impact of the fixed factors Bound and Verb Type on the availability of scalar readings with TEO predicates in contrast to TCoS predicates, as outlined in (33).

Following the considerations aforementioned, the (salient) bound in the scale structure of the verb is expected to have an impact on the availability of a sentence to yield a lower- or upper- bounded reading with both TEO and TCoS predicates to the effect that those items that present a minimum standard degree elicit a lower-bounded inference, and those that specify a maximum standard degree yield an upper-bounded inference. Note that the expectation for TEO predicates is related to where the realization (or threshold) of the mental state is attained, proposed differently for particular groups of verbs: the onset of the state for inchoative state causative verbs (e.g., deleitar ‘delight’, cf. Fábregas and Marín 2015; Fritz-Huechante et al. 2020), and the culminating endpoint of a process for psych accomplishment verbs (e.g., serenar ‘calm down’, cf. Alexiadou and Iordăchioaia 2014; Landau 2010, among others).

5.2 Material

The factor Bound was implemented by 12 verbs per Verb Type (see Appendix: Table 5 for a list of TEO verbs, and Table 6 for a list of TCoS verbs). For TEO verbs, the items were partially chosen from a larger inventory elicited by a questionnaire targeting alternating psych verbs featuring basic emotion domains, i.e., anger, disgust, fear, happiness, and sadness (see Rott et al. 2020; Rott and Verhoeven 2019). Within the upper-bounded TEO verb class, which constitutes a reduced group similar in meaning, items formed with the prefix des- were selected depending on the meaning of the morphologically basic item, the requirement being to be a base with a negative meaning. This was in order to avoid an interpretation in terms of politeness (cf. Horn 1989), and hence obtain a lower-bounded reading of the item. TCoS verbs were chosen from the existing literature for both lexical types, lower- and upper-bounded (cf. Hay et al. 1999; Kennedy and Levin 2008; Levin 1993; McKoon and Macfarland 2000; Rappaport Hovav 2014; Wright 2002, among others). We further screened both TEO and TCoS verbs with respect to frequency considering both macro-varieties of the language, Peninsular Spanish and Latin American Spanish. Remember that an inanimate causer entails a change of state in the object (cf. Alexiadou and Anagnostopoulou 2020; Fritz-Huechante et al. 2020; Martin 2020) which is necessary for this study, hence only those items that allow for the accommodation of an inanimate causer in the sentence structure were selected. The experiencer objects were all proper names. In the case of the objects in TCoS sentences, first these were selected according to their frequency in relation to the verbal item at hand. In addition, avoiding specific proportional specifications (e.g., la hoja de papel ‘the sheet of paper’) was also taking into consideration in order to prevent a strong incremental theme reading that could aid to a mismatch between lower-bounded predicates and upper-bounded readings. Hence, objects were presented as general as possible (e.g., el papel ‘the paper’) and keeping in mind the naturalness of the sentence.

In order to test the availability of scalar readings, we used sentences of the type introduced in (31) and (32). Sentences (34) and (35) illustrate the four conditions of the experiment. Verbs were embedded under the structure of (34) and (35) and sentences were presented individually to the participants. After each sentence a question targeting the state of the affected object was displayed: Is x in the state? (see (34b), (34d), (35b), (35d)). Participants were instructed to choose the best answer to the question between two options: (a) para nada ‘not at all’, and (b) bastante ‘quite a lot’.

Expectations for upper-bounded predicates (cf. (34)) are that participants choose the answer bastante ‘quite a lot’. This is because the adverb casi ‘almost’ selects the values that are near and adjacent to the maximum degree of the relevant scale. Hence, TEO sentence in (34a) is interpreted as Camila was calmed down to a degree close to total calmness caused by the melody, and TCoS (34c) is interpreted as the clothes were dry to a degree near completeness caused by the wind. In the case of lower-bounded predicates (cf. (35)), we expect participants to choose the answer para nada ‘not at all’ since almost yields readings that a change of state in the object does not occur with these predicates. The interpretation of sentence TEO (35a) is that Camila did not start the state of being delighted caused by the story, and TCoS (35c) is that the paper did not change into the state of being wrinkled caused by the printer. Table 2 gives an overview of the expected effects of the study.

Based on a Latin-square design, we created 4 lists, each containing 12 items (6 items of each Verb Type). Items also represented the two different values of the factor Bound (lower-bounded and upper-bounded), so that each list had three repetitions of each experimental condition (i.e., 3 lower-bounded and 3 upper-bounded TEO verbs, 3 lower-bounded and 3 upper-bounded TCoS verbs). No items were repeated. The target verbs were mixed with 36 fillers that also presented a decision between para nada ‘not at all’ and bastante ‘quite a lot’ responses. Further, fillers presented four levels of difficulty, the first level used as a control level to check the concentration of the participants and accuracy on the task at hand (accuracy threshold: 75 %). All sentences were automatically randomized by participant. Each participant saw only one list. All sentences were presented in the simple past (pretérito indefinido).

5.3 Procedure

Seventy-six Spanish native speakers (19 per list) participated in the experiment. Three were excluded due to not reaching the accuracy threshold of 75 % in the control level set. A total of 73 participants ranging in age from 15 to 58 (36 females, 35 males, 2 no gender, age M = 35.88) were considered for the analysis. Participants were monolinguals living mainly in Chile and some other Latin American countries (Argentina, Colombia, Costa Rica, Ecuador, El Salvador, Honduras, Mexico, Peru) as well as Spain. All subjects took part of the study voluntarily and anonymously. The study included a 5-practice trial which was presented prior to the beginning of the experiment to familiarize each participant with the task. The experiments were conducted as web-based studies and implemented on Ibex Farm HU. Each session took approximately 20 min and was unpaid.

5.4 Results

The obtained data consists of outcomes of a binary variable: para nada ‘not at all’ and bastante ‘quite a lot’. There were no missing values, i.e., (73 speakers × 4 conditions × 3 answers =) 876 answers were collected. Table 3 presents the proportions of these outcomes broken down by the experimental factors Verb Type and Bound (see Appendix for the proportion of outcomes per verb, Table 5 for TEO verbs and Table 6 for TCoS verbs).

Figure 1 visualizes the results of Table 3. Confidence intervals (at a 0.95 level) were calculated with a binomial test (assuming a two-sided hypothesis; calculations were made with the function binom.test of the package stats, 4.0.3, in R). Inspection of the results points to an effect of Bound on the frequency of para nada ‘not at all’ responses: they were chosen more often with lower-bounded verbs, either TCoS verbs (79.91 %) or TEO verbs (77.63 %) than upper-bounded verbs (TCoS: 47.95 %, TEO: 64.38 %). Moreover, the data shows a larger decrease in the choice of para nada ‘not at all’ for upper-bounded TCoS verbs compared to lower-bounded TCoS verbs, whereas the difference between upper-bounded and lower-bounded TEO verbs is smaller.

Figure 1:

Proportions of para nada ‘not at all’ responses by Verb Type and Bound.

The data was fitted with a generalized linear mixed-effects model in R (R 2013, Version 3.0.2). The fixed factors are Verb Type (level 0: TCoS; level 1: TEO) and Bound (level 0: lower; level 1: upper). Subjects and Items were modeled as random factors. The maximal random-effects structure involved intercepts of Subjects and Items as well as slopes of Subjects with Verb Type and Bound; slopes of Items with Verb Type and Bound cannot be evaluated since items (lexicalizations of different verbs) are nested in the fixed factors. The maximal random-effect structure does not converge, which means that it cannot be evaluated with the present dataset. The models with slopes of Subjects with either fixed factor converge, while the model with a slope of Subjects with Verb Type reaches a better AIC value (845.7) than a model with a slope of Subjects with Bound (856.7). This is the most parsimonious random-effect structure; it was kept maximal during the statistic evaluation of the fixed factors (cf. Barr et al. 2013).

The fixed effects were tested with a backwards-elimination procedure, based on model comparison with Likelihood Ratio tests. The model of maximal fit is reported in Table 4. The interaction between Verb Type and Bound did not reach significance (Likelihood Ratio Test: χ² = 2.56, p = 0.11). In order to estimate the significance of the main effects, a model with two main effects was compared with a model in which the effect of interest was removed. The results in Table 4 show that both factors have a significant impact on the response variable.