PhiN 75/2016: 49
Richard Waltereit (Newcastle)
J'ai lu ce livre moi-même: On the synchrony and diachrony of French intensifiers
Intensifiers have attracted a great deal of attention
recently, especially where English and German are concerned (cf. König and Siemund 2000a, Siemund 2000, Gast 2006, Eckardt 2006). While they are not reflexive
pronouns, they are in a close historical and typological relationship with
them. Whereas intensifiers in English and German have been studied quite
extensively, there is very little research on French intensifiers (see
Zribi-Hertz 1995 though). This article will study properties of French
intensifiers in a synchronic as well as a diachronic perspective.
1 Intensifiers in English
1.1 Intensifiers vs. reflexives
Intensifiers in English are the forms himself/herself like in (1)–(5).
They are associated with an NP in the same clause: In (1), herself refers to the same entity as the rector; in (3), himself refers to the same entity as Tony Blair. They share this characteristic with reflexive pronouns, as exemplified in the following sentence:
However, there is an important difference to reflexive pronouns: whereas the latter represent an argument on their own (or at least a slot of what would be an argument in the transitive counterpart of the reflexive), the former do not. In (6), the NP the author and himself represent different arguments (subject and NP-argument, respectively). By contrast, in (1)–(5), the NP and the self-form are associated with the same argument slot.
PhiN 75/2016: 50
1.2 Types of intensifiers
Siemund (2000), as well as König and Siemund in a number of jointly authored papers (e.g. 2000a, 2000b), distinguish three types of intensifiers.
The first type is the "adnominal" intensifier, like in (1). Adnominal intensifiers modify an NP. They contrast the NP against a set of potential alternative referents, where the target NP and the set of potential alternatives are in a centre-periphery relation, with the target NP in the centre. The centre-periphery structure is hierarchical: thus, (1) contrasts the Rector with an implied less central figure; (2) contrasts the Queen with her entourage.
The second type of intensifiers, according to Siemund, are the "adverbal-exclusive" ones, like in (3). They modify a VP, rather than an NP. They can often paraphrased as 'alone', 'on their own'. Their contribution to meaning is that the subject of the verb phrase in question did the action "with their own hands", rather than delegating it to someone else or having it done by proxy.
The third type, according to König and Siemund, are "adverbal-inclusive" intensifiers like in (4). According to Siemund, they modify the VP, just as the adverbal-exclusive ones. Also, just as the adverbal-exclusive ones, they are associated with the subject of the verb phrase in question. They can be paraphrased by 'also', ‘as well’ and convey that the NP they are associated with is included in a set of potential referents. Thus, (4) includes the speaker in an implied set of people who have made the same experiences; (5) includes the speaker in an implied set of reader of that book. Adverbal-inclusive are furthermore characterized by rich contextual inferences that make the contribution made by this type of intensifier difficult to appreciate outside context. For example, (4) evokes a context of perhaps difficult personal experiences; in (5), the sentence I have read it myself, without any context, would probably intuitively evoke the adverbal-exclusive reading. It’s only the context given in the example that makes the adverbal-inclusive reading plausible.
It may be asked whether these three uses are really distinct functions of the form, or merely contextual effects. Thus, Rooryck & Vanden Wyngaerd (2010: 59n) claim, when referring to the various contrastive uses of the intensifier, that "this is a function of the pragmatic context, and as such the effect can easily be lifted by changing the context". However, Siemund (2000: 12) offers the following argument: Intensifiers can co-occur in the same proposition, as in (7) (Siemund's example):
PhiN 75/2016: 51
In this unrealistically artificial yet perfectly acceptable example, himself is used as adnominal, adverbal-exclusive, and adverbal-inclusive intensifier, respectively, meaning that 'Bill, in contrast to somebody else, had to be told the answer and that he was not the only one for whom that was necessary' (Siemund 2000: 12). The acceptability of such constructions is strong support for the notion that the three uses are indeed separate meanings of the form, rather than being merely contextually induced. Even stronger support for that notion comes from their diachronically staggered availability in French, as we will see later in section 5.5.
Next, I would like to investigate whether French has similar types of intensifiers.
2 Intensifiers in French
There is some information on French in the "Typological Database on Intensifiers and Reflexives" (TDIR, Gast et al. 2007). According to the source, French has all three types of intensifiers, expressed by lui-même / elle-même. I will now study them in turn.
In French, lui-même / elle-même can be used as an adnominal intensifier.
PhiN 75/2016: 52
As in English, this use of the intensifier evokes a centre-periphery scenario with NP lui/elle-même referring to a "centre", contrasting with an appropriate "periphery", for example a king and its entourage, a product and the scenario of its purchase, or a city and its conurbation. As Siemund (2000: 136–153) notes, this centre-periphery arrangement invokes genuine encyclopaedic knowledge, rather than discourse structure. In other words, the speaker is not entirely free to add adnominal lui-même/elle-même to any NP they wish to highlight as a centre. Consider this minor adjustment of (11):
While it is a perfectly reasonable thing to say that it is more common to hear English than Spanish in Miami's suburbs and nothing should prevent the speaker to focus on the suburbs rather than the city, it is odd to mark this with elle-même. This is because in the city/suburbia frame set up by ville, the ville is, quite literally, the centre.
Next, let us look at the adverbal-exclusive intensifier use of lui-même/elle-même. Firstly, some examples.
The intensifier, while associated with an NP, really modifies the VP. It sets up a contrast in absentia between the predicate and another more general one that is ambiguous as to whether the subject performed the action in person or delegated it. Indeed, construire sa maison is ambiguous insofar as it can be said of someone who has their house built by builders, as well as of someone who lays the bricks of their house with their own hands. Adding lui-même / elle-même, however, will exclude the former option and solely permit the latter one. With trancher, though, that ambiguity is more difficult to obtain. Indeed, the contribution of the intensifier is to focus on the verb, rather than setting up a contrast with a situation where that activity would be delegated. Thus, the contribution made by the adverbal-exclusive intensifier is at the level of predicate meaning.
PhiN 75/2016: 53
Finally, let us move on to adverbal-inclusive intensifiers in French. I start by giving some examples.
As with the other uses of the intensifier, the form lui-même/elle-même is associated with an NP. As noted with the English examples above, in consonance with Siemund (2000)'s observations, the adverbal-inclusive intensifier "includes" the NP in question in a set of referents or potential referents: in (14), the subject is a railway worker like others are; in (15), the subject has lost weight like someone else could, for example the addressee. As such, they can be replaced with 'also' or other additive markers. But this is not everything. It is striking that in this type of usage, the form is used in contexts of argumentation. This is quite obvious in the examples cited: the fact that the subject is a railway worker gives their claim about SNCF's safety record additional credibility. Likewise, the fact that in (15), the subject has lost weight themselves adds to the credibility of the claim that they can help the addressee lose weight. In short, the adverbal-inclusive invokes the experience associated with the predicate in question. This leads us to take a fresh look at the distinction between adverbal-exclusive and adverbal-inclusive intensifier.
3 On the nature of the distinction between adverbal-exclusive and adverbal-inclusive intensifiers
3.1 Siemund (2000): Distinction in verbal semantics
The labels "adverbal-exclusive" and "adverbal-inclusive" used by Siemund as well as by König and Siemund suggest that the distinction is a binary contrast. Indeed, Siemund attempts to capture the distinction by proposing a number of minimal contrasts that decide between an adverbal-exclusive and an adverbal-inclusive reading of the intensifier. These contrasts are located at a broadly aspectual level, but at any rate at the level of predicate meaning. The proposed key contrast is between transferable and repeatable situations. Transferable situations are characteristic of adverbal-exclusive intensifiers, whereas repeatable situations are typical of adverbal-inclusive situations. In fact, a non-repeatable situation, according to Siemund, automatically triggers the adverbal-exclusive reading. A situation that is both transferable and repeatable, however, would be compatible with both the exclusive and the inclusive reading of the intensifier. "Transferable" essentially means that the same event could apply, or could have applied, to a different subject as well. "Repeatable" means that the situation is not unique, that it can occur more than once. Thus, in
PhiN 75/2016: 54
moi-même could be read either as exclusive or as inclusive adverbal intensifier. The situation could apply to a different subject (the speaker could ask someone else to read the book aloud, as opposed to reading it with their own eyes), thus yielding an adverbal-exclusive reading; or the situation could be repeated (the book can be read by various people), thus yielding an adverbal-inclusive reading.
As Siemund notes, adverbal-exclusive and adverbal-inclusive have somewhat different preferences when it comes to situation types and definiteness of the objects (in the case of transitive predicates), without however blocking the respective opposite. The adverbal-exclusive prefers achievements and accomplishments, as well as definite direct objects; the adverbal-inclusive, by contrast, prefers states and processes, as well as indefinite direct objects. However, if anything, this is really only a preference, rather than a robust generalization. Consider, for example, (16): we are in business with an accomplishment and a definite direct object here, yet the sentence is perfectly compatible with an adverbal-inclusive reading. Conversely, states or processes may be compatible with the adverbal-exclusive reading:
This would be a perfectly acceptable utterance in, for example, a context where there is an expectation for people to be fed rather than taking in food with their own hands. The adverbal-exclusive is, equally, compatible with indefinite objects:
Thus, the aspectual criteria proffered cannot be really decisive, and would be, if anything, the by-product of a distinction at a more fundamental level.
Siemund makes it clear that the adverbal-inclusive is conducive to generating pragmatic inferences, in particular conditional ones, and often used to introduce a request. However, by his own admission (2000: 228–229), he is unable to offer a genuine explanation for this.
Looking at the criteria "transferable" and "repeatable" as used to characterize the adverbal-exclusive and adverbal-inclusive intensifier, respectively, it would appear that they essentially paraphrase what is inherent in the informal characterisation of these usage types already. Thus, "transferable" merely paraphrases the notion that a particular action can be delegated to someone else, which is already inherent in the descriptive characterization of the adverbal-exclusive type. Likewise, the definition Siemund (2000: 185) offers of "repeatable" merely spells out the descriptive characterisation of the adverbal-inclusive type of intensifier usage: "Non-repeatable situations immediately trigger AVS [adverbal-exclusive intensifier]. […] Situations restricted to happening only once contradict the very idea of inclusion."
The conclusion that seems to impose itself here is that the distinction between adverbal-exclusive and adverbal-inclusive cannot really be located at the level of situation semantics. Rather, the basis for this distinction needs to be sought elsewhere.
PhiN 75/2016: 55
3.2 Adverbal-exclusive vs. adverbal-inclusive: content-level vs. context-level
A concept that promises to be relevant for the contrast between adverbal-exclusive and adverbal-inclusive intensifiers is the content-level vs. context-level distinction (Hansen 2008). It refers to the truth-conditional vs. non-truth-conditional use of polysemous items. Some examples follow.
(19) exemplifies a well-known polysemy. In (19), the function of because is a different one in (19a) than in (19b): In (19a), the fact of getting a job in Brussels is the reason why Claire moves there. By contrast, in (19b), the fact that there is a blue Golf in front of the house is not the reason why Tom is in; rather, it is the reason for the speaker to say that Tom is at home. Thus, whereas in (19a), the conjunction because establishes a material link between two statements about the world and is thus truth-conditional, the same conjunction in (19b) does not establish a material link between two statements about the world. Rather, in (19b), the conjunction because relates a belief of the speaker's to the world. In other words, the truth of (19a) can be evaluated against facts in the real world, whereas the truth of (19b) cannot. As Hansen (2008: 14) explains, the reason for this is that because in (19a) has a "content-level" reading whereas in (19b) it has a "context-level" one. The same distinction applies to the adverb toujours in (20). In (20a), toujours refers to a phase in a given state of affairs, and its truth can be evaluated against this state of affairs. It is a content-level use of the phasal adverb (Hansen 2008: 15). In (20b), by contrast, it does not characterise a state-of-affairs; rather, it describe an attitude of the speaker. This is a context-level use of the same adverb (Hansen 2008: 15).
The contrast content-level vs. context-level does not, in itself, reflect a specific syntactic distinction. In other words, while a context-level use is likely to be located at a higher level of syntactic structure than a content-level one, there is no specific syntactic level associated with either of the two (cf. Hansen 2008: 16).
Nor does the contrast content-level vs. context-level match a particular level of semantic representation. That is, both content-level and context-level uses could be conventionalized functions of the same lexical item, i.e., meanings of a polysemous item; just as well, however, could a context-level use of some form can be merely a contextually induced variant of its content-level counterpart, rather than being a fully-fledged conventional meaning. Thus, it may be argued that the use of toujours in (20b) is a conventional context-level one, because it would not be possible to get the same effect with counterparts of toujours in other languages, e.g. English. By contrast, the use of because in (19b) may merely be a contextually induced context-level variant. After all, the context-level use of a causal conjunction is common with such conjunctions in many languages (e.g. Aijmer 1997).
PhiN 75/2016: 56
Returning to the distinction between adverbal-exclusive and adverbal-inclusive intensifiers, it would appear that the difference between the two can be captured by assigning the former to content-level and the latter to context-level. The adverbal-exclusive intensifier has a truth-conditional function; after all, it adds a specific element to predicate meaning. It specifies that the subject did the action with their own hands/eyes, rather than leaving it open whether the event was carried out by the subject themselves or by proxy. This specific element may be checked against the instance of a state of affairs – it is truth-conditional. By way of example, the sentences of the minimal pair in (21) may have, quite obviously, different truth conditions.
In particular, if by (21a) it is conveyed that the subject has the house built, rather than building it herself, the sentence is not true under the same circumstances as (21b). Syntactically, the adverbal-exclusive intensifier modifies the verb phrase (VP).
The adverbal-inclusive intensifier, by contrast, has a different function. It invokes the experience of the referent associated with the intensifier, in order to give greater credibility to a claim the speaker is making. It presents its host utterance as an argument for a conclusion. As such, it does not affect the truth conditions of the sentence. Compare the clauses of this minimal pair:
The two sentences do not differ in truth conditions. Syntactially, the function of the adverbal-inclusive intensifier is at clause, rather than at verb phrase level. It is located above the verb phrase level (Gast 2006). As a consequence, it is not actually adverbal in the syntactic sense. Nonetheless, I will keep with the established terminology and continue to refer to them as adverbal intensifiers. The distinction between adverbal-inclusive and adverbal-exclusive intensifier may, at times, be reflected in word order. Compare the two variants of (23):
(23a) is ambiguous. It permits the adverbal-exclusive reading 'I have read this book (with my own eyes, rather than having it read out aloud by someone else)' as well as the adverbal-inclusive reading 'I have read this book (I know what it is about)'. By contrast, (23b) only allows the adverbal-inclusive reading. The way word order interferes with intensifier reading is complex, though. It should not be inferred, for example, from the contrast in (23) that the position between finite verb and participle as in (23b) generally blocks the adverbal-exclusive reading. Thus, (24)
PhiN 75/2016: 57
would be compatible with both the adverbal-exclusive and the adverbal-inclusive reading.
The preceding discussion on the nature of the distinction between the two adverbal intensifiers puts us in a position to address the relationship between the three types of intensifiers at the level of semantic relations.
4 On the relationship between the three types of intensifiers
As the three types of intensifiers are functions of the same form, we need to ask what the semantic relation between these functions is. Semantic relations between readings of the same item generally are of four types: metaphor, metonymy, hyperonymy or hyponymy (cf. Blank 1997, Gévaudan 2007). Metaphor is based on a perceptive or functional similarity. Metonymy is based on a focus/background shift in a frame (cf. Koch 2004). Hyperonymy and hyponymy are converse relations of taxonomic super- or subordination, respectively. The following polysemies exemplify the four relations.
It is irrelevant for the purposes of this classification whether the readings are conventional meanings of the form, as arguably with examples (25) and (27), or merely contextual effects, as possibly in (26).
While all of these are relevant in lexical meaning, Detges & Waltereit (2002: 164–165) have shown that only metonymy and hyperonymy/hyponoymy are available in grammatical meaning. As far as intensifiers are concerned, I contend that the relation between the three is consistently metonymic. This can be seen in appropriate "bridging contexts", i.e. contexts that are compatible with both among a pair of readings under consideration. I will now turn to this.
Firstly, let us look at the relation between adnominal and adverbal-exclusive intensifiers.
PhiN 75/2016: 58
It is not entirely clear whether we are dealing with an adnominal or an adverbal-exclusive intensifier here. Is it the king himself (as opposed to someone in his entourage) who took the notebooks, or is the point that the king took the notebooks himself, i.e., not delegating this to someone else? This makes the link between the two functions evident: if someone of high rank (as opposed to someone in their entourage) does something, it is reasonable to infer that the "high-rank individual" did it with their own hands. This is a metonymic link – the two readings are closely related through inference but neither perceptually similar nor inclusive in a logical way. Note, though, that even though the inference looks natural, it is not automatic: it would be perfectly possible for the king "to take the notebooks" by ordering one of his servants to do so. Thus, the two readings are clearly distinct. The former does not entail the latter even though the inference seems very natural. Note, equally, that the shift from adnominal to adverbal intensifier implies a syntactic rearrangement. With the adnominal reading, the intensifier adjoins to the NP, whereas with the adverbal reading, it adjoins to the VP.
Turning now to the link between adverbal-exclusive and adverbal-inclusive intensifier, the connection between them is equally metonymic. This may be appreciated in example (29).
In this example, an adverbal-exclusive intensifier is put to use as an argument for a conclusion: the fact the electrician had built his own house himself is presented as evidence for the claim that he is a knowledgeable adviser for house-hunting. This can be presented as evidence because doing something with one's own hands, for example building a house, means that one has gained the relevant experience, and that one can reliably count as an expert in the matter. This is the metonymic link between the two readings. Contexts like (29) make the link between adverbal-exclusive and adverbal-inclusive palpable. They are closely related in an encyclopaedic way, i.e. through knowledge of the (extra-linguistic) world.
PhiN 75/2016: 59
The metonymic link implies a reanalysis of the intensifier from content-level to context-level. In the adverbal-inclusive reading, the intensifier does not modify the VP; by the same token, it does not add to verb meaning.
In sum, there are metonymic links between the adnominal and the adverbal-exclusive, as well as between the adverbal-exclusive and the adverbal-inclusive reading.
After having identified, in the abstract, metonymic links between the intensifier readings, it is worthwhile examining whether these match the historical build-up of polysemy in the history of the intensifier.
4.1 Diachronic profile of intensifier readings
We now move to a diachronic study of the intensifier in the history of the French language. The goal here is not to provide an exhaustive historical analysis of the intensifier, but rather to show that the availability of the various functions is historically staggered. The data are from the Textes de Français Ancien (TFA) and the FRANTEXT corpora. I only examine the form lui-même, not même.
There are very few occurrences of intensifiers in the Old French period. The very first example is an adnominal one.
Alongside other clearly adnominal ones, we next find some examples that are ambiguous between an adnominal and an adverbal-exclusive reading. One has already been discussed in the preceding subsection, repeated here for convenience.
Below is another example exhibiting the same kind of ambiguity.
PhiN 75/2016: 60
There is some ambiguity here: is it the "goddess Venus herself" (as opposed to, say, mortal beings) that is picking the flowers, or is the intended meaning that Venus picked the rosary herself (rather than having it done by someone else)? The former alternative would mean that we are dealing with an adnominal intensifier, the latter however would imply that we are in business with an adverbal-exclusive one.
These two examples confirm is that there is a metonymic link between the adnominal and the adverbal-exclusive intensifier: a high-ranking individual doing something invites the inference that the individual did this on their own, rather than delegating it to someone else.
Later still, we find unambiguously adverbal-exclusive intensifer uses of lui-même/elle-même:
The following example of an adverbal-exclusive intensifier from the early 17th century shows how its meaning can be put to use as an argument in discourse to support a conclusion.
The fact that the narrator has seen themselves their interlocutor "following someone in the woods" can be invoked to support the claim that the interlocutor committed a treachery. This is essentially the same argumentative pattern as the one used in (29): In the same way as having built one's house with one's own hands can be presented as an argument for being knowledgeable in all things houses, seeing with one's own eyes that a person followed someone where they weren't supposed to can be used as an argument to support the conclusion that that person committed a treachery. Having done / experienced something personally can be invoked as an argument for being a credible authority in the matter. In a similar fashion, the mere mention of "having done something personally" allows the inference of suitable relevant experience.
PhiN 75/2016: 61
Later on, we find the adverbial-inclusive use, where the previous inference of relevant experience is conventionalized as a separate semantic function of the intensifier. One example is (34).
The use of lui-même here is clearly an adverbal-inclusive one. Abraham's being wealthy "himself" is presented as not consistent with Abraham's own condemnation of wealth. Thus, rather than operating at content-level, the function of lui-même is located at the level of expressing the narrator's attitudes – the context-level. It could not be read as an adverbal-exclusive use. After all, it makes no sense to say of someone that they are rich by themselves. Being wealthy is a property that an individual can only have at a personal level – it cannot be delegated to someone else. If one hands over one's wealth to someone else, then it becomes someone else's wealth. If one gave someone else power of attorney, this would not make the relevant assets property of the attorney, and it would not affect the individual in question's personal wealth.
It would be conceivable, though, to think of this intensifier as an adnominal one, assuming that Abraham is a person of high rank. However, this seems unlikely. For one, the intensifier is placed next to the verb, quite distant from the NP. This is indicative of adverbal status. Moreover, in the context of the utterance, what is important is that Abraham's wealth is contrasted with his own condemnation of it, rather than contrasting Abraham with any person in his entourage. This again suggests an analysis as adverbal-inclusive intensifier.
The diachronically profile of intensifiers in French is, furthermore, important from a theoretical point of view. Firstly, it offers additional support for the assumption that the intensifier readings are distinct conventional meanings rather than contextual effects. Co-occurrence of the three readings in the same sentence provided already some support for this assumption (cf. (7)). Diachronic staggering of the readings makes the assumption even stronger. Secondly, it confirms Siemund's (2000: 180) implicational hierarchy of intensifier functions. Siemund submitted that if an intensifier is used as adverbal-inclusive, then it is also used as adverbal-exclusive, and that if an intensifier is used as adverbal-exclusive (but not necessarily as adverbal-inclusive), then it is also used as adnominal. This is confirmed for French.
PhiN 75/2016: 62
In this artcile, intensifiers in French were studied. French lui-même / elle-même has, like its English counterpart himself/herself, three main functions: adnominal, adverbal-exclusive, and adverbal-inclusive. I discussed Siemund's (2000) analysis of the distinction between adverbal-exclusive and adverbal-inclusive. While the distinction itself was found to be highly insightful, I suggested that its nature cannot be located in propositional meaning alone. Rather, I suggested that the contrast content-level vs. context-level, proposed by Hansen (2008), is a more promising way of analysing the contexts of use of this pair.
I also analysed the diachrony of the intensifier in the French language. Use of the adnominal intensifier is attested from Old French. The adverbal-exclusive is found from the 16th century onwards, and the adverbal-inclusive from the 17th century onwards. This confirms Siemund's (2000) implicational hierarchy of intensifier functions.
FRANTEXT = [http://www.frantext.fr]
BFM = Base de français médiéval [http://bfm.ens-lyon.fr/]
Blank, Andreas (1997): Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen [Beihefte zur Zeitschrift für Romanische Philologie 285]. Tübingen: Niemeyer.
Detges, Ulrich and Richard Waltereit (2002): Reanalysis vs. Grammaticalization: A semantic-pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21: 151–195.
Eckardt, Regine (2006): Meaning Change in Grammaticalization: An enquiry into Semantic Reanalysis. Oxford: Oxford University Press.
PhiN 75/2016: 63
Gast, Volker (2006): The Grammar of Identity: Intensifiers and reflexives in Germanic languages. London: Routledge.
Gévaudan, Paul (2007): Typologie des lexikalischen Wandels. Bedeutungswandel, Wortbildung und Entlehnung am Beispiel der romanischen Sprachen. Tübingen: Stauffenburg.
Hansen, Maj-Britt Mosegaard (2008): Particles at the Semantics/pragmatics Interface: synchronic and diachronic issues. A study with special reference to the French phasal adverbs. Bingley: Emerald.
König, Ekkehard & Siemund, Peter (2000a): The Development of Complex Reflexives and Intensifiers in English. Diachronica 17: 39–84.
König, Ekkehard & Siemund, Peter (2000b): Intensifiers and Reflexives: A typological perspective. In Reflexives: Form and function, Zygmunt Frajzyngier, Traci S. Curl (eds) [Typological studies in language 40], 41–74. Amsterdam and Philadelphia: Benjamins.
Rooryck, Johan & Vanden Wyngaerd, Guido (2010): From First Principles. Anaphors without Binding Theory. Brussels: Hogeschool – Universeit Brussel.
Siemund, Peter (2000): Intensifiers in English and German. A comparison. London: Routledge.
Zribi-Hertz, Anne (1995): Emphatic or Reflexive? On the endophoric character of French lui-même and similar complex pronouns. Journal of Linguistics 31: 333–374.
1 This contrast was pointed out to me by Alain Berrendonner.