Exaptation , Refunctionalization , Decapitalization — BE + Past Participle with Intransitive Verbs in Mediaeval and Early Modern Spanish

The chapter presents the current state of research concerning the development of the BE + past participle constructions from Latin to Spanish. Starting from the description in Rosemeyer (2014) and the theoretical background collated in Kailuweit and Rosemeyer (2015), it will be shown that the functional change does not follow traditional grammaticalization paths. Several concepts that deal with cases contradicting traditional grammaticalization theory will be discussed. ‘Exaption’ (Lass 1990, 1997), focusing on total defunctionalization does not account for the fact that the resultative value of the BE + past participle construction, marginal in Latin, becomes central in Mediaeval Spanish. ‘Refunctionalization’ Smith (2008, 2011) captures this aspect in a more appropriate way. However, the development of the construction could be also conceived as the opposite of what Pountain (2000) describes as ‘capitalization’: a process of ‘decapitalization,’ by which a feature is exploited, not for wider, but for more restricted purposes.


Introduction
The formation of compound tenses in Romance languages with the periphrasis HAVE + past participle has been considered a prototypical case of grammaticalization (Schwenter 1994;Squartini and Bertinetto 2000;Detges 2001), following a grammaticalization path that seems to be universal (Bybee et al. 1994).In line with Harris (1982), Detges (2001) pointed out that four stages have to be distinguished.At the first stage, the value of the construction is resultative, marking the present state resulting from past action.At the second stage, the construction marks a continuous or repeated series of events that started in the past.At the third stage, the construction refers to a past event with present relevance (anteriors in Bybee et al.'s 1994 terminology) and at the fourth stage, it attains an aorist function denoting a past event without present relevance.While in standard Spanish HABER + past participle has only reached the third stage, AVOIR + past participle in spoken French has gone further on to the fourth stage, taking over the function of the simple past that is excluded from oral registers.
As far as the development of BE + past participle from Latin to Romance languages is concerned, the situation is quite complex.At first glance, there is no change at all.The structure BE + past participle already exists in classical Latin, expressing perfect passive (amatus est) and the perfect of the so-called deponent verbs (locutus est, profectus est) (Vincent 1982, pp. 85-86).According to Vincent (ibid.,p. 86), these verbs favored "the development of periphrastic expressions like *ven(u)tum est," because etymologically they go back to medio-passives in Indo-European and, therefore, do not assign the role of a prototypical agent to their subject.Many deponent verbs denote a change of state that the subject-argument undergoes (gradior 'I step', morior 'I die', nascor 'I am born ', etc.).Hence, a more or less formal and functional continuity from Latin to Romance, as far as BE + past participle is concerned, has been claimed in the literature (Yllera 1980;Vincent 1982;Harris 1982;Aranovich 2003).
However, Jacob (1994, pp. 363-65) pointed out that passive construction with BE + past participle in Romance languages has a presence-orientated resultative function incompatible with the preterite value of the formally analogous Latin periphrastic passive.According to Jacob (ibid.,p. 365) and Mackenzie (2006, pp. 132-35) the same holds for deponent verbs.Jacob (ibid.,p. 363) highlighted the fact that a regression from preterite to resultative would contradict all common grammaticalization theories. 1 The objective of this short chapter is basically theoretical.I will cast some new light on the development of BE + past participle from Latin to Early Modern Spanish.I will do so by discussing some data I extracted from the CORDE-corpus, 2 but essentially on the basis of the studies of Rosemeyer (2014Rosemeyer ( , 2015) ) and, regarding Latin and Proto-Romance data, Cennamo (1997Cennamo ( , 2001Cennamo ( , 2008)).My central claim is to take up the main topic of the present volume.Given a particular counter-example to the unidirectionality of a grammaticalization path, I shall discuss conceptual alternatives of description, for example, exaptation (Lass 1990(Lass , 1997;;Narrog 2007), refunctionalization (Smith 2011) and capitalization (Pountain 2000).

BE + Past Participle-A Criterion for Unaccusativity?
In the last few decades, the fact that many modern European languages exhibit an alternation between the auxiliaries HAVE and BE in compound tenses of intransitive verbs has become a central topic of linguistic research.Starting with Perlmutter (1978) Unaccusative Hypothesis, 3 auxiliary selection has been discussed as the main criterion for the distinction of an unaccusative (verbs selecting BE) and an unergative class (verbs selecting HAVE) (Alexiadou et al. 2004;Aranovich 2007;Kailuweit and Rosemeyer 2015).Although BE-selection is not a necessary criterion for unaccusativity, it is deemed a sufficient one.If we consider, in line with Sorace (2000) and Legendre and Sorace (2003), that unaccusativity is gradient, 4 it is possible that BE-selection situates an intransitive verb nearer to the unaccusative pole (see Table 1).
BE-selection characterizes unaccusative core verbs, not only in modern Italian and French, but presumably in all languages using the BE/HAVE opposition to mark split-intransitivity in compound tenses.Having lost the use of BE + past participle in their standard varieties, Ibero-Romance languages seem to reflect the distribution of ÊTRE/AVOIR we find in contemporary standard French (Kailuweit 2011(Kailuweit , 2015) ) at the mediaeval and early modern stage.The most prototypical verbs occurring with BE + past participle are verbs of change of location and verbs of change of state, while verbs of controlled processes select exclusively HAVE. 5 The following examples from Medieval Spanish may illustrate this contrast: (1) The spread of HAVE in the history of Spanish seems to follow Sorace (2000) auxiliary selection hierarchy (Benzing 1931;Aranovich 2003;Mateu 2009;Rosemeyer 2014).However, the claim that earlier stages of Ibero-Romance languages code an opposition of unaccusativity and unergativity via auxiliary selection is based on the assumption that the BE + past participle and the HAVE + past participle construction with intransitive verbs are functionally equivalent, that is, instances of verbal perfect (anterior).As Mackenzie (2006) and Rosemeyer (2014) have shown, this assumption is by no means evident.

ESSE + Past Participle < SER + Past Participle
In classical Latin, the periphrastic perfect of deponent verbs-as well as the perfect in general-has the main function of indicating the accomplished process, even if the result of this process is still given: 6 Le perfectum [ . . .] correspond en gros à la valeur du parfait grec indiquant le résultat acquis d'un procès et à celle de l'aorist grec ; il recouvre à la fois l'un et l'autre, sans avoir la valeur propre de chacun des deux : son rôle est d'indiquer le procès accompli.(Meillet and Vendryes 1924, p. 248) In (3) the absolute construction referring to advanced age indicates that the perfect mortua est denotes the process of dying and not the fact of being dead.
( The same type of construction appears in Early Modern Spanish.In (4) the de-PP refers to the cause that operates in the process of dying ('to die from').On the contrary, the por-PPs in the Mediaeval Spanish example (5) indicate the reasons for being dead ('to be dead because of').
( These examples may illustrate Mackenzie (2006) approach, worked out and proven through a large corpora analysis by Rosemeyer (2014).Mackenzie (2006, pp. 131-39) put forward the hypothesis 6 See Ramos Guerreira (2009) for a current overview of the functions of the Latin Perfect.The basic assumptions of Meillet and Vendryes (1924) are still valid.7 Rosemeyer (2014) argues convincingly that early Modern Spanish begins around 1425.Hence, the objection of an anonymous reviewer that example (4) may have been copied from an older version stemming from the 16th century does not affect the argumentation.that, firstly, the perfect of Latin deponent verbs underwent "syntactic reanalysis" 8 (ibid., p. 133) and that secondly, a "cyclical pattern of development" (ibid., p. 139) occurred that converted the copula + verbal adjective9 (again) into a perfect (see Table 2).
He justified his claim by arguing that in the Proto-Romance period, synthetic perfect forms were created (sequi, mori, nasci, etc.) even for those verbs which "exhibit a deponent-style E-perfect" 10  (ibid., p. 133). 11This process led to a "generalization of the synthetic perfect" on the one hand and, on the other hand, a "reorganization of the membership of the class of intransitives that entered into the periphrastic construction" (ibid., pp. 133-34).Latin deponent verbs with an activity aktionsart, such as fabulari ('to speak'), were ruled out, and presumably prototypical unaccusatives such as venire ('to come') 12 were included (ibid., p. 134).According to Mackenzie (2006), the maintenance of the periphrastic construction with a semantically more homogeneous group is due to a functional split.The synthetic perfect functions as an anterior while the periphrastic construction is reanalyzed as a copula + verbal adjective construction with a "resultant-state" type of meaning.Mackenzie (ibid., conceded that there are examples of BE + past participle from Mediaeval Romance languages that suggest an anterior reading.However, most occurrences should be analyzed as resultative constructions.According to Mackenzie (ibid.,p. 136), the resultative reading is supported by the distribution of perfect and imperfect forms with an aspectual, rather than anterior, meaning due to the temporal context.Finally, the copular analysis explains the "otherwise surprising omission of the reflexive pronoun" (ibid., p. 137) in constructions such as es levantado that reads rather 'is on his feet' and not 'has got up'.
Some additional criteria should be taken into consideration.In their analysis of the decline of the BE-perfect construction in Canadian French, Sankoff and Thibault (1977, pp. 85-94) proposed a number of criteria to distinguish BE + past participle as an anterior and as a copula construction.Adverbials referring to the manner in which an event is realized are evidence against the copula + adjective interpretation.Adverbials expressing the motivation of a subject referent for an action favor an eventive interpretation.Finally, if the construction does not denote a resultant state that persists at reference time, it has to be interpreted as an anterior.Mackenzie (2006) only illustrated his findings with a couple of examples.Hence, his claim that the "vast majority of the examples in the mediaeval corpus" exhibit a resultant-state type of meaning remains intuitive.Rosemeyer (2014) has applied the mentioned criteria systematically to vast corpora of Spanish data of the mediaeval and early modern period.Following Detges (2001), he asserts that there is a metonymic link between anteriors and resultatives (ibid., p. 49): a resultant state has been brought about by a previous event and a previous event may lead to a resultant state that persists at reference time.Therefore, the interpretation is ambiguous in the absence of any of the markers discussed in Sankoff and Thibault (1977) and Mackenzie (2006).Nonetheless, meticulous interpretation and statistical analysis led Rosemeyer (2014, p. 37) to the claim that, "in its prototypical use, Old Spanish ser + PtcP does not express anteriority."In addition, Rosemeyer proved that in Early Modern Spanish, the opposition between SER + past participle and HABER + past participle "was reanalyzed as an opposition between two auxiliaries for one and the same construction, the anterior" (ibid., p. 255).Due to the metonymic relation of anteriors and resultatives, the SER + past participle-construction underwent a functional change, probably supported by the spread of HABER + past participle to less prototypical verb classes in terms of Sorace (2000) Auxiliary Selection Hierarchy (see Table 1).The growing restriction of SER-selection to high frequency verbs of change of location (entrenchment) led to the formulation of a new rule: when confronted with the choice between SER + past participle and HABER + past participle, use SER + past participle if you wish to express a past change of location event (ibid., p. 234).In Kailuweit (2015), I claimed that an analogous process took place in French between the 17th and 19th century, leading to considerable loss of the ÊTRE + past participle construction with verbs of change of state and a fossilization of the ÊTRE + past participle construction with verbs of change of location.As in Early Modern Spanish, the remaining cases of BE + past participle are not restricted to resultative readings. 14n light of the theoretical questions discussed in this volume, the details of this second reanalysis may not be as pertinent as those of the first.While the second reanalysis follows the grammaticalization path for perfects observed in an extensive number of languages, the fact that the opposite process seems to have taken place from Latin to Romance is puzzling.

Discontinuity
Being aware of the striking contradiction to established grammaticalization theories, Jacob (1994) denied any continuity from Latin deponent perfects to the Old Spanish SER + past participle construction.He pointed out that the resultative construction does not need any historical motivation since it consists of the combination of the copula BE and an adjectivized past participle (ibid., p. 363). 15owever, it remains unclear why the association of the copular BE + past participle construction with the formally identical perfect construction of deponent verbs was lost.
In line with Jacob (1994), Cennamo (2008, p. 125) rejected the hypothesis of a formal and functional continuity between analytic perfects of Latin deponents and Romance BE past participle constructions.Nonetheless, as Cennamo (ibid.)pointed out, the discontinuity is not an effect of the loss of deponents, as one might suggest, but, on the contrary, it is due to generalization of deponent perfect structure in Late Latin. 16According to Cennamo (ibid.,p. 121), a central change took place in the transition from Latin to Romance as far as the encoding of transitivity, grammatical voice and relation was concerned.The data that Cennamo (2008) compiled show puzzling equivalences and interchangeability among voice forms.Hence, the active may be used in passive function and the passive may occur in active function.Therefore, the passive no longer consistently marked the undergoer17 argument and the active no longer marked the actor argument in subject function.A form, such as amatus sum in ( 6 Hence, Jacob (1994, p. 363) hypothesis of an always-accessible copular interpretation of be falls short of a satisfactory description.As the data in Cennamo (2008) show, in contexts where BE + past participle continues to be associated with perfective readings, a copular analysis is excluded.
An additional problem came up with the loss of the case system.As long as case-marking the nominative could be distinguished from the accusative, speakers could easily identify the semantic roles of verbal arguments.Hence, the loss of the voice system illustrated in (6) did not substantially impair communication.Nonetheless, case-marking started to operate on an active-inactive basis (Cennamo 2008, p. 122).The accusative appeared in subject function, to mark the undergoer argument of intransitives, such as anticausatives and other verbs denoting change of state or location, passives and impersonals.Accusatives could be found inclusively with non-agentive activities and even mark the actor of transitive verbs, although the nominative form continued to be the canonical case for subjects.Hence, case-marking no longer helped to consistently identify the semantic roles of verbal arguments.A construction showing the prior form of the passive voice such as puellam amata(m) est (7) out of context could be ambiguous between an active interpretation with puellam marking the actor (7a) or, with two arguments, the undergoer (7b).It could also have a passive interpretation, with puellam marking the undergoer (7c-d): The interaction of the two phenomena (the loss of the voice opposition and case-marking indicating syntactic and semantic status), illustrated by puellam in (7), leads to significant changes taking place in the encoding of the argument structure of the clause in the transition from Latin to Romance: the generalization of the synthetic perfect, the grammaticalization of HABERE + past participle marking the Actor function of the subject with transitive and, in the long run, also intransitive constructions and, last but not least, the reanalysis of BE + past participle marking the undergoer argument of a resultative construction with intransitive verbs, especially those denoting change of state (Mackenzie 2006;Cennamo 2008;Rosemeyer 2014).
Note that the findings contradict not only the continuity of a perfect use of ESSE + past participle to SER + past participle but also the continuity of a copula function of esse in combination with the verbal adjective that formally continues the past participle.The well documented cases of active readings of forms of the amatus sum type (Cennamo 1997(Cennamo , 2001(Cennamo , 2008) )  18 seem to indicate that values attributed to ESSE + past participle in Classical Latin have become defunctionalized. 19Table 3 resumes the findings of the last two sections of this chapter:

Exaptation, Refunctionalization, Decapitalization
As we have seen, the development of Latin ESSE + past participle to Mediaeval Spanish SER + past participle contradicts the common assumption of a grammaticalization path from resultative constructions to aorist constructions.Actually, the development seems to be more or less the other way round.While the analytic perfect of Latin deponent verbs had an aorist and an anterior reading and appear to be only marginally ambiguous allowing a resultative reading, Mediaeval Spanish SER + past participle constructions are clearly not aoristic and have to be interpreted in most cases as resultative constructions.An anterior reading only marginally comes into play.However, this development does not seem to be a motivated reversal of the grammaticalization process but rather a reset after a period of defunctionalization.
In the following, I will briefly discuss three concepts brought up in the last years in the context of questioning the irreversibility und unidirectionality of grammaticalization paths: exaptation, refunctionalization and capitalization.My objective will be to show that the development of BE + past 18 An anonymous reviewer criticised that my argumentation depends too much on the works of Cennamo and the references she adduces.However, as Latinists confirmed to me, her references are valid and even more current studies of Latin Grammar, such as Baños Baños (2009), do not contradict her findings.Since my claim in this paper is primarily a theoretical one, this is not the place to add wider bibliographic and empirical support to confirm her findings.See Drinka (2017, pp. 139-40) for a recent affirmative summary of Cennamo (2008). 19I use the term 'defunctionalization' in line with Smith (2011, p. 269) as "the loss of value or function of an opposition."participle from Latin to Spanish, sketched out in the last two sections of this chapter, is not an idiosyncratic case but a specific type of counter-example to grammaticalization theory that could be described with the help of the aforementioned concepts.I shall, however, invert the concept of capitalization introduced by Pountain (2000) and discuss a process that I call 'decapitalization'.
The concept of 'exaptation' was introduced by Lass (1990Lass ( , 1997) ) as a limited alternative to grammaticalization in specific functional domains.Exaptation makes use of the general concept of reanalysis that I have already used in the contribution and will not discuss on its own.20In Lass' terms, exaptation, adapted from biology where it "denote[s] the co-optation during evolution of structures originally developed for other purposes"21 is "a kind of conceptual renovation, as it were, of material that is already there but either serving some other purpose, or serving no purpose" (Lass 1997, p. 316).As Narrog (2007) shows, Lass' examples for linguistic exaptation, the re-use of Indo-European ablaut inter alia, are not always convincing and even if they are, they could be easily considered isolated counter examples to the unidirectionality of grammaticalization.However, Narrog (2007) himself proves that causativization at several stages of Japanese as well as in a couple of unrelated languages occur in a systematic way that contradicts the claim of grammaticalization theory and could be better described in terms of exaptation.
In his 1990 paper, Lass highlighted the possibility that languages recycle morphological material that does not serve to code a functional distinction anymore and therefore could be considered 'junk'.According to Cennamo (2008) findings, this seems to be precisely the case for ESSE + past participle in Late Latin.Apparently, both functional values coded by this construction in Classical Latin were jettisoned; the passive function (amatus sum opposed to amavi) as well as the perfect function (amatus sum opposed to amor).But instead of being dropped, the form persisted and was recycled as a present tense passive marker for transitive22 and a resultative marker for intransitive verbs.
Nonetheless, the concept of morphological junk has been criticized in the literature.23As Narrog (2007, p. 7) pointed out: "it is rarely the case that linguistic structure is completely functionless or useless."Lass himself conceded in 1997, that "the most common kinds of exaptation [ . . . ] are not based on co-optation of junk.'Useful' [ . . .] features can be adapted too" (Lass 1997, p. 318).
In this line of thinking, we could account for the fact that the construction ESSE + past participle-even if it allows for an active transitive reading at some point in history-is not completely functionless.The conjugated form of esse still codes the categories of person and number. 24Presumably, speakers recognized these categories easier in the case of esse as an irregular verb than in the case of regular conjugated full verbs affected by phonetical abrasion on their endings.
In a more recent paper, Smith (2011) took up the discussion on exaptation on the bases of Lass (1997).He pointed out that there is a conceptually significant distinction between cases of exaptation including loss of the original function and others with a new function grafted onto the original one (Smith 2011, p. 305).Following Smith, the first case should be labelled refunctionalization, the second adfunctionalization.Since the main function of ESSE + past participle in Classical Latin, the perfect function including an aorist reading, is lost leading to defunctionalization, the recycled use of the forms in a resultative construction could be described as an instance of refunctionalization in Smith's terms.
However, according to Smith (2011), refunctionalization as the acquisition of a new value or function by a morphological opposition, does not recycle morphological 'junk' but equals a process that has been described in the history of art and architecture as "skeuomorphy" (ibid., pp. 275-76).A 'skeuomorph' is a "feature which starts off as functional, loses its functionality and becomes decorative-that is, in some sense, cultural (ibid., p. 277)."They still have content, that is, they are more in tune to people's expectations, so objects or buildings designed with skeuomorphs would be more aesthetically pleasing that those without skeuomorphs.Applied to the case of refunctionalization, skeuomorphy explains a vestigial continuity between the old and the new function: Whilst the original functional distinctions disappear, the formal oppositions that used to express them seem to retain a vestige of abstract content-they become, in the terminology introduced earlier, skeuomorphy, not junk-and, where one of the items is discarded or the opposition assumes a new function, it is this vestigial content that seems to determine what developments take place.(Smith 2011, p. 314) If we follow Smith's argumentation, our description of the development of ESSE + past participle in Late Latin has to be revisited.It is not an instance of total defunctionalization.It seems to be that the marginal resultative function of the Classical Latin ESSE + past participle construction is maintained as a vestige.I suppose that it is in line with the findings of Cennamo (2008) to say that 'resultativity' attains the more abstract value of affect or concern.After the loss of the perfect and the passive value of the structure, this abstract vestigial content explains its refunctionalization in a new organized system of argument marking, where ESSE + past perfect highlights the undergoer function of the argument in contexts of change of state and change of location.
Let me conclude this section by bringing one last concept into play: capitalization introduced by Pountain (2000) in order to describe processes of weak grammaticalization without phonetical reduction: Use of the term 'capitalization' is an attempt to label the historical process by which a linguistic feature which already exists in a language comes to be substantially exploited for wider purposes, sometimes simply making overt distinctions which were previously covert but sometimes apparently creating new expressive possibilities.(Pountain 2000, p. 295) I would like to raise the question of whether Pountain's considerations could be useful for describing a somehow inverted process that, consequently, has to be labeled 'decapitalization'.If we substitute 'wider purposes' with 'tighter purposes', the first part of Pountain's definition seems to fit well to our case.While in Classical Latin, ESSE + past perfect had a (dominant) perfect and aorist value and appeared only marginally as a resultative, in Mediaeval Spanish, SER + past participle is mainly a resultative construction.Hence, it has been decapitalized.The process of decapitalization turns overt distinctions into covert ones.The overt perfect value of ESSE + past participle in Classical Latin was (almost) covert in Mediaeval Spanish and appeared only marginally in the data.The apparently new expressive possibility of marking an undergoer argument became central.However, mutatis mutandis, the function of marking an undergoer continued the semantic properties not only of passives but also of deponent verbs, although in a more systematic way.
What would be the added value of a concept labeled 'decapitalization' in the theoretical field discussed in this volume?While 'capitalization' seems to operate in the same direction as grammaticalization, 'decapitalization' works counter-directional to grammaticalization.As we have seen, the fairly well-established concept of exaptation refers to a wide range of counter-directional processes.According to Smith (2011), at least two forms of exaptation have to be distinguished: refunctionalization and adfunctionalization.As far as refunctionalization in Smith's terms is concerned, the defining criterion is 'skeuomorphy', a highly technical concept of art history that, in my opinion, is not completely convincing in the field of linguistics.In art history, skeuomorphy refers to elements that are maintained as decorative elements, although having lost their original functions.Smith (2011, p. 277) himself conceded that 'decorative,' "is not a term usually associated with serious studies of language." 25However, it seems to be more problematic to substitute, as Smith does, 'decorative values' in art history with 'a vestige of abstract content' in linguistics.
My claim is that we could better deal with this type of linguistic change if we consider the vestigial abstract value as a remnant of a process of decapitalization.Let me discuss this concept briefly.My considerations are, of course, inspired by Pountain's concept of capitalization.However, it seems to be fruitful to have a look at the use of 'decapitalization' in economics and social science.According to Padley (1986, p. 183), 'decapitalization' means "the loss of capital for accumulation" or as Van Brabant (1987, p. 388) puts it: "the stock of capital shrinks over time."In my opinion, it would be more enlightening to use this concept metaphorically in historical linguistics than the opaque term 'skeuomorphy', especially in cases of a decrease in structural distinctions.Other grammatical changes from Latin to Romance might be described as decapitalization, too; for example, the loss of the case system.The morphemes -a(m), -u(m) and -as, -os decapitalized their functional load.While denoting case, number and gender in Latin, they only refer to number and gender in Romance languages.But, of course, this is only speculative suggestion, which has to be discussed more in detail in an additional study.

Conclusions
In this chapter, I based my argument on the studies of Cennamo (1997Cennamo ( , 2001Cennamo ( , 2008) ) and Rosemeyer (2014), in order to show that the development of BE + past participle from Classical Latin to Mediaeval and Early Modern Spanish neither exhibits functional continuity nor equals the development of HAVE + past participle.While HAVE + past participle embarks upon a grammaticalization path that is held universally, BE + past participle seems to take the opposite path at the first stage, regressing from an aoristic construction in Classical Latin to a resultative construction in Mediaeval Spanish.Since this development contradicts with common insights on grammaticalization, the continuity of the construction from Latin to Spanish has been denied.Exaptation theory (Lass 1990(Lass , 1997) ) seems to be a theoretical framework that deals with processes of linguistic change that exhibit only continuity on the formal side, while on the functional side, it comes to total defunctionalization.The morphemes left-in our case the BE + past participle structure-appear to be morphological 'junk' that might be recycled in order to adopt new functions.However, functional discontinuity does not seem to be complete in many cases.As Smith (2008) pointed out, it is more probable that when functional opposition is jettisoned, a vestigial abstract content remains on the marked element.In our case, it is the resultative value, marginal for Classical Latin ESSE + past participle that is maintained, presumably as a more abstract feature of affect or concern.This abstract feature determined refunctionalization by paving the way to a new functional load for SER + past participle in Mediaeval Spanish: the marking of an undergoer subject.The fact that the perfect and aorist functions of ESSE + past participle are jettisoned while the marginal resultative value becomes central in Mediaeval Spanish, as well as in other Mediaeval Romance varieties, should be described as decapitalization.The concept of 'decapitalization,' with its connotations stemming from economics and social science, denotes that in a given situation of linguistic change, a feature could be exploited for tighter purposes.In the development from Latin to Proto-Romance, this seems to happen not only to ESSE + past participle but also to the nominal endings expressing gender, number and case.Further research will have to clarify to what extent the concept of 'decapitalization' is useful for describing processes of functional reduction in language change, not only in the special case of the development of Proto-Romance, but also for a wider set of unrelated languages.