6.1.1 The relation stereotype from Martin & Piñón (2016)

As various previous authors have suggested, the relevant part of the corresponding noun meaning in a behavior-related verb is the typical – and by “typical” we now mean stereotypical or prototypical – properties associated with the noun meaning, though we will speak of stereotypical properties (i.e. stereotypes) and assume that prototypical properties are among them.¹¹ The initial idea is to postulate a relation stereotype between nominal properties N and stereotypes S, as in (30), such that S is a stereotype (i.e. a stereotypical property) of N. For example, if N were diplomat, then stereotype(S, diplomat) would state that S is a stereotype of diplomat (i.e. of diplomats), for example, being discreet or carrying a nice briefcase or being well-groomed.

In (30), we assume for simplicity a classical, non-kind treatment of noun meanings as predicates of individuals, but our approach could be recast using kinds if desired. The relation stereotype could also be intensionalized in various ways, for example, it could be made world-dependent and/or context-dependent, but this is not crucial for our present purposes. Note that although stereotype is an undefined relation in our approach, the hope is that this relation is needed independently of our analysis of behavior-related verbs.¹² Even so, there are three principles that apply to stereotype as we conceive of this relation. The first principle says that every stereotype S of a nominal property N is a property of an individual x or a property of a state s or a property of an event e:

The second principle says that if S is a stereotype of N, then S does not entail N, which is to say that S is not a hyponym of N:

Finally, the third principle informally states that if S is a stereotype of N, then S is based on the “facts” of individuals that are N. More formally and verbosely, this principle says that if S is a stereotype of N, then there is an x such that N applies to x, and either S applies to x, or there is a relation R such that R is a thematic relation, and either there is a state s such that S applies to s and R holds between s and x, or there is an event e such that S applies to e and R holds between e and x. Another way of saying this is that this principle requires S to be “grounded” in an individual x that N applies to in such a way that either S applies to x or S applies to a state that x participates in or S applies to an event that x participates in.

In view of this third principle, it will be useful to define a relation exhibit between individuals x and properties S and N (“x exhibits S with respect to N”) such that S is a stereotype of N, and either S applies to x, or there is a relation R such that R is a thematic relation, and either there is a state s such that S applies to s and R holds between s and x, or there is an event e such that S applies to e and R holds between e and x:

6.1.2 Deriving verbs from common nouns

After these remarks on the relation stereotype, let’s turn to the question of how behavior-related verbs are derived, using diplomatiser ‘behave like a diplomat’ as an example. The noun diplomate ‘diplomat’ is straightforwardly analyzed as the following predicate of individuals:

Applying the relation stereotype in (30) to this predicate, we derive the set of stereotypes S of diplomats:

In (36), the stereotypes S of diplomats may be properties of individuals, states, or events (recall (31)).

Instead of postulating that the meaning of -iser applies to a predicate of stereotypes such as the one in (36), it seems more natural to think of -is- in the suffix -is-er as spelling out a verbal head in charge of introducing stereotypes, for otherwise it would not be clear what element introduces them into the derivation. This verbal head, which we call v_stereo, can be spelled out either by -is- (or -ifi-), or by a covert null suffix in the case of other behavior-related verbs like patienter.

It is not the case, however, that all -iser-verbs involve the stereotype relation. For instance, a predicate such as alcooliser ‘put alcohol in a liquid, cause to get drunk’ arguably does not involve stereotypicality. Therefore, v_stereo should not be equated with -is- itself; rather -is- may spell out v_stereo (as in diplomatiser), or not (as in alcooliser).

Furthermore, recall that most -iser behavior-related verbs additionally have (anti)causative uses, that also involve the stereotype relation, see e.g., (3)). Consequently, we ideally would like our analysis of the verbal head v_stereo spelled out by -is- in the -iser suffix to be extendable to these additional uses. Therefore, we factor out the agentive flavour of Martin & Piñón’s (2016) analysis for -iser. More concretely, we do not make the verbal head v_stereo it realizes responsible for introducing an external argument, and leave the eventuality v it introduces underspecified between events and states. This motivates the following analysis of v_stereo as spelled out by -is-, which we label as “-is-_stereo”:

The predicate in (37) applies to a nominal predicate P, and an eventuality v, and yields the conditions that there is a property S such that S is a stereotype of P, S holds of v, and v is an event or a state. Applying this relation to the nominal predicate in (35), we obtain the following analysis of diplomatiser_stereo:

To obtain the unergative use, the eventuality predicate in (38) is combined with a Voice head (Kratzer 1996) that introduces an external argument x of an eventuality v that is an event:

Applying (39) to (38), we obtain the unergative predicate (40), where the alternative that v is a state is eliminated.

Applied to an individual x and an eventuality v, this relation yields the conditions that x is the agent of v, and there is a property S such that S is a stereotype of diplomats and S holds of v and v is an event.

We can now apply the relation in (40) to the individual constant juliette (for Juliette), and derive the predicate of eventualities (more precisely, of events) in (41).

6.1.3 The figurative reading of nouns

We now turn to the use of a noun on its figurative reading in an indefinite NP, for example, recall (7), repeated below.

Since in the figurative reading, the relevant part of the noun is again the stereotypical properties associated to it, we can again employ the relation stereotype to treat this reading. However, in this case, it is arguably the meaning of the indefinite article un/une ‘a(n)’ that introduces stereotypes. We use the subscript “f” when un_f/une_f applies to a noun and introduces stereotypes, thus triggering the selection of the figurative use of the noun. There are two readily available ways of analyzing un/une ‘a(n)’ as introducing stereotypes.

The first is give un/une ‘a(n)’ a non-quantificational (predicative) analysis: the meaning of the indefinite article¹³ applies to a nominal property N and yields a predicate of individuals x such that there exists a stereotype S that x exhibits with respect to N, as in (42).

Applied to the predicate diplomat, this meaning of the indefinite article yields the following predicate, which denotes the set of individuals x such that there is a stereotype S x exhibits with respect to diplomats:

The following simple-minded analysis of the copula est ‘is’ (ignoring tense) takes the copula to apply to a predicate P of individuals in order to yield a relation between states s and individuals x such that P applies to x and x is the theme of s:

Observe that the states s denoted by this analysis of est ‘is’ are “light” in that they barely have any descriptive content: the only condition is that the individuals x are their themes. This “lightness” suggests the following innocent principle, which says that if an individual x stands in a thematic relation R to an event e, then there is a state s such that x is the theme of s:

In other words, if an individual x participates (thematically) in an event e, then there is a state s that x is the theme of. This principle will be useful below.

Applying the analysis of est ‘is’ in (44) to the predicate in (43), we derive the relation between states s and individuals x such that there is a stereotype S that x exhibits with respect to diplomats and x is the theme of s:

If this relation is applied to the individual constant juliette, we obtain the predicate of states s such that there is a stereotype S of diplomats that Juliette exhibits and Juliette is the theme of s:

We will briefly mention the second readily available way of analyzing un/une ‘a(n)’ as introducing stereotypes, which is a quantificational analysis (hence un/une_{f_q}):

In (48), R is a relation (corresponding to the VP meaning) between eventualities (events or states) v and individuals x. Applied to the predicate diplomat, the following quantifier is derived, which if applied to a relation R between eventualities and individuals, yields a predicate of eventualities v such that there is an individual x and a stereotype S such that x exhibits S with respect to diplomats and R holds between v and x:

This use of un/une_{f_q} diplomate ‘a diplomat’, which is figurative and quantificational, figures in sentences such as (10).

We conclude this section with the remark that the analyses presented above allow us to account for why the sentence in (9) with diplomatiser ‘behave like a diplomat’ (see (41)) entails the sentence in (9c) with une_{f_nq} diplomate ‘a diplomat’ (see (46)), ignoring tense. This entailment is due to the following fact:

The proof of this fact is straightforward and uses the definition in (34) and the principle in (45). Intuitively, this entailment is valid because a stereotype that makes (9) true is necessarily eventive, but then it counts as a stereotype that also makes (9c) true.

We can also show that the reverse entailment is not valid (recall also (14)), because a stereotype that makes (9c) true need not be eventive, whereas a stereotype that makes (9) true is necessarily eventive.

6.1.4 Deriving verbs from proper nouns

Unsurprisingly, we adopt the same basic approach to behavior-related verbs from proper nouns, but with the difference that the stereotypes are now of individuals as opposed to nominal properties (sets of individuals). Note, however, that the relation stereotype as given in (30) is not applicable to individuals directly, and so we need to define a derived relation, designated by stereotype′, between stereotypes S and individuals x, which effectively treats x as a singleton (the set of individuals identical to x), as shown in (51).

As an illustration, let’s consider the proper noun Sarkozy and its standard treatment as an individual constant, here sarkozy:

Applying the relation stereotype′ to this constant, we obtain the set of stereotypes of Sarkozy:

In order to derive the behavior-related verb sarkozyser ‘behave like Sarkozy’, we need a version of -is- (cf. -is-_stereo in (37)) that is applicable to individuals instead of nominal properties. This version, -is-_stereo′, is analogous to -is-_stereo but makes use of stereotype′ in place of stereotype:

The behavior-related verb sarkozyser ‘behave like Sarkozy’ is then derived via the application of -is-_stereo′ to Sarkozy (triggering elision of the /i/) and the combination of the input of this operation with the Voice head (39):

As seen in (55), the result is a relation between events v and individuals x such that x is the agent of v, v is an event, and there is an S such that S is a stereotype of Sarkozy and S applies to v.

Applied to the individual constant macron, the relation in (55) yields (ignoring tense) the predicate of events e such that Macron is the agent of e, e is an event, and there is an S such that S is a stereotype of Sarkozy and S applies to e:

The present approach can be naturally extended to treat examples where a proper noun appears with an indefinite article:

To treat the figurative use of the indefinite NP in (57), it is first convenient to define a derived relation exhibit′ that is based on the relation stereotype′, analogous to the relation exhibit from (34) (which is based on stereotype):

The next step is to propose an analogue of un/une_{f_nq} from (42) that makes use of the relation exhibit′, applying to individuals:

Applied to sarkozy, this relation yields the predicate of individuals x such that there is a stereotype S that x exhibits with respect to Sarkozy:

If the meaning of est ‘is’ given in (44) is then applied to this predicate, the following relation between states and individuals is derived (cf. (46)):

Finally, if we apply this relation to the individual constant macron, we obtain the predicate of states s such that there is a stereotype S that Macron exhibits with respect to Sarkozy and Macron is the theme of s, which is arguably a reasonable rendering of un (vraie/autre) Sarkozy ‘a (true)/another Sarkozy’ in (57) (neglecting the adjective):

We point out that on this approach, Macron sarkozyse_pn ‘Macron behaves like Sarkozy’ entails the sentence in (57), but not vice versa, for the same reasons as before (cf. (50)):

6.2.1 Semantic analysis

We now extend further the analysis to deadjectival behavior-related verbs, using bêtifier ‘behave in a stupid way’ as an example.

The adjective bête ‘stupid’ is analyzed as a predicate of individuals:

The adjective itself does not introduce stereotypes. So for instance, (65) simply states that Juliette is stupid, not that she has stereotypical properties of stupid people.

Just like with nouns, it is the suffix (or the indefinite un_f/une_f when the adjective can be nominalized) which is in charge of introducing the stereotypical relation. Turning to the case of bêtifier, we give the verbalizing suffix -ifi-_stereo the same analysis as for -is-_stereo (see (37)):

Applying the predicate -ifi- to the adjectival predicate in (64) and combining the resulting predicate with the Voice head (39), we obtain the following analysis for bêtifier (again, where the alternative that v is a state is eliminated).

Applied to an individual x and an eventuality v, this relation yields the conditions that x is the agent of v, and there is a property S such that S is a stereotype of stupid individuals and S holds of v and v is an event. We can apply the relation in (67) to the individual constant juliette (for Juliette), and derive the predicate of eventualities (more precisely, of events) in (68).

6.2.2 Deriving the entailment patterns between the verb and the adjective

As illustrated in (19), the sentence with the deadjectival behavior-related unergative verb does not entail the sentence with the corresponding adjective (ignoring tense), see also the absence of entailment between (69a) and (69b) below.

(69)	a.	Juliette bêtifie.
		‘Juliette behaves as stupid people typically do.’
	b.	Juliette est bête.
		‘Juliette is stupid.’

The entailment does not succeed for sentence (69a) only requires Juliette to perform an act e such that a stereotypical property of stupid individuals holds of this act e, what she can do without being stupid herself.

We also explain why sentences with a behavior-related verb derived from an adjective P do not entail that the events they denote satisfy the property denoted by P, as required by the manner adverbial derived from the adjective (remember that sentences such as Juliette patiente impatiemment ‘Juliette is waiting impatiently’ are not contradictory). This is so because an event e may satisfy a stereotypical property of the set of individuals defined by the adjective (e.g., the set of stupid entities) without e being in that set (e.g., without being a stupid event); recall our second principle (32).

We also account as before for why the sentence with the adjective (e.g., (69b)) does not entail the sentence with the corresponding behavior-related verb (e.g., (69a)). While the verbal sentence necessarily ascribes an eventive property, the adjectival sentence may also ascribe a stative one.

It happens, however, that the adjectival sentence indicates – through, e.g., the perfective aspect – that the property ascribed is eventive (see, e.g., Jean a été patient). Even in this case, the entailment towards the corresponding behavior-related verb is blocked, because the behavior-related verb, but not the adjective, requires a stereotype to hold of e.

6.3.1 Morphosyntactic restrictions

A number of behavior-related verbs undergo the causative alternation, and thus have causative and anticausative uses in addition to their unergative use. This is rather surprising, given that unergatives typically do not participate in the causative alternation in languages such as French or English (see (70a)); in fact, causativization of unergatives is even more restricted in French than in English, since in the latter language, causativizing unergatives is possible in resultative constructions (which do not exist in French); see the contrast between (70b) and its English translation.¹⁴

The availability of the additional causative and anticausative uses depends on the morphological make-up of the verb. Most behavior-related unergative verbs formed with the suffix -is- or -ifi- can be used as causative or anticausative verbs, even if these latter uses are sometimes not lexicographed in dictionaries (as is the causative use of diplomatiser, which is not listed in the TLFi), see for example, (3) repeated below.¹⁵

In contrast, “zero-derived” behavior-related verbs in general do not undergo the causative alternation; see the generalization 1 in (71).

For instance, the examples in (72), all found on the Internet, become infelicitous when the suffixed variant is replaced by the zero-derived variant enclosed in square brackets.

Other examples of behavior-related -iser or -ifier with salient causative and anticausative uses are given in the Appendix in (102).

An apparent exception to the generalization 1 in (71) is the zero-derived verb materner ‘mother’, sharing its root with the adjective maternel ‘maternal/motherly’. This verb has a transitive use, as in (73).

However, while (3) entails that the theme has a typical property of diplomats, (73) entails that the agent behaves (with the theme) as mothers typically do. Furthermore, the transitive use of (73) is not causative. In Rappaport Hovav & Levin’s (1998) typology, materner under its transitive use should rather be classified among so-called non-core transitive verbs, which are manner (activity) verbs. One of the diagnostics differentiating causatives (which are core transitive verbs) from non-core transitives is whether the direct object can be omitted (and this without the verb enduring a meaning shift). As (73) shows, it can in the case of materner. In contrast, in (3), the omission of the object would force the selection of the unergative meaning (under which Sarkozy rather than Hezbollah has a typical property of diplomats), which is not entailed by (3). As observed by Rappaport Hovav & Levin (1998), Alexiadou et al. (2015) and others, core vs. non-core transitive verbs differ by their aspectual properties and event structure; while the former have a complex event structure – an event and a result state in Alexiadou et al.’s (2015) framework – and yield telic predicates when combined with a quantized object, the latter denote processes and may also yield atelic predicates when combined with a quantized object. This difference is reflected in the distribution of frame adverbials, see (74).

Finally, like non-core transitive verbs, materner does not anticausativize (it cannot be used to mean become motherly), and transitive sentences built with such verbs do not entail a change in the theme’s referent.

In summary, predicates such as materner are not true exceptions to the generalization (71) which says that zero-derived behavior-related verbs do not undergo the causative alternation, since they do not have a causative semantics in any of their uses.

We call predicates such as materner behavior-related non-core transitive verbs. This subtype of behavior-related verbs is poorly represented in French and seems much more frequent in English, see the list in (75), from Clark & Clark (1979). The English predicates in (75) have no French counterpart; tyranniser ‘tyrannize’ and vamper ‘vamp’ are among the few French examples of this subclass of behavior-related verbs.

A second generalization on the way morphosyntax constrains the range of meanings behavior-related may have concerns prefixed verbs, see (76).

For instance, bêtifier ‘(cause to) behave in a stupid way’, niaiser ‘behave in a simple-minded way’, canailler ‘behave as a rascal’ all are behavior-related unergative verbs, see (77a) and (78a), but the prefixed verbs derived from the same adjectival or nominal root – abêtir ‘make (more) stupid’, enniaiser ‘make (more) stupid’, encanailler ‘strip somebody of their social status by making them socializing with scoundrels’ – do not: the latter only have causative or anticausative uses, see (78) and (80). Table 1 offers a summary of the restrictions imposed by the morphosyntactic makeup of behavior-related verbs on their unergative, (anti-)causative and non-core transitive uses.¹⁶

6.3.2 Semantic analysis

In order to capture the causative and anticausative uses, we start again with (38):

In the causative sentence (3), v is a state of which the referent of the internal argument is the theme. In the spirit of recent work by Zeller (2001) and Lohndal (2014), we severe the internal argument from the verbal predicate and assume that this argument is introduced by a separate head. We attribute the semantics (81) to this head (that we label as ‘Trans’, after Zeller 2001).

Supposing that the verbal predicate diplomatiser ‘diplomatize’ is combined with the verbal head Trans (81), we obtain the following semantic analysis for this VP:

Applied to an individual y and to an eventuality v, this relation yields the conditions that y is the theme of v, and there is a property S such that S is a stereotype of diplomats and S holds of v, and v is an event or a state.

We can now apply the relation in (82) to the individual constant hezbollah (for Hezbollah), and derive the predicate of eventualities in (83).

In the next step, the VP represented in (83) is combined with a head introducing an eventuality v″ causing a result v″′ which is a state. We call this head ‘Cause’, see (84).¹⁷

Applying (84) to (83), we obtain the following predicate of events, where the alternative that v is an event is eliminated:

The analysis of diplomatiser in (85) captures the anticausative use, exemplified in (4) repeated below, where se is semantically inert; see Schäfer (2008) for the syntactic details.¹⁸

In order to obtain the causative use of this predicate, we combine (85) with the Voice head (39) (Alexiadou et al. 2006):

The predicate in (86) applies to an individual x, and an eventuality v″ such that x is the agent of v″, v″ is an event, v″ causes a v, and there is a property S such that S is a stereotype of diplomats, S holds of v, Hezbollah is the theme of v, and v is a state.

Finally, we extend the analysis to non-core transitive behavior-related verbs, using tyranniser ‘tyrannise’ as an example. The strategy consists in deriving first a transitive predicate via the application of Trans, and to combine the resulting predicate with the Voice head. The head Cause does not enter the composition of tyranniser, which is not causative when used transitively, as observed in section 6.3.1.

We start with the noun tyran ‘tyrant’, analyzed as a predicate of individuals, see (87), and derive the set of stereotypes S of tyrants by applying the relation stereotype in (30) to this predicate, see (88). Next, we apply the predicate -is-_stereo to the resulting nominal predicate, and obtain the predicate in (89).

In the next step, we combine (89) with the Trans head (81), and obtain (90).

Next, the predicate (90) with Juliette for the value of y is combined with the Voice head (39) introducing the external argument, which yields the relation in (91), where the alternative that v is a state is eliminated:

Applied to an individual x and an eventuality v′, this relation yields the conditions that x is the agent of v′, Juliette is the theme of v′, v′ is an event, there is a property S such that S is a stereotype of tyrants, and S holds of v′. Thus this relation does not convey any change in Juliette, which conforms to the observation in section 6.3.1.

Finally, let us see how the analysis can account for the restrictions imposed by the morphosyntactic makeup of the verb on the range of available uses displayed by behavior-related verbs, recall Table 1 and the generalizations (71) and (76) repeated below.

To account for (71), we simply stipulate that while both -is-/-ifi- and the null suffix may spell out the verbal head v_stereo, only -is-/-ifi- may optionally also spell out the Cause head necessary to obtain the causative and anticausative predicate (without necessarily doing so however, since some behavior-related verbs built with -is-, such as tyranniser, do not have a causative meaning).

To account for (76), we assume that the prefix necessarily spells out Cause and Trans, making the predicate transitive and causative, thereby blocking the unergative use. This is in line with previous findings by Junker (1987), showing that prefixes tend to act as transitivizers in French.