Noun Phrases as expressions of evidentiality : an analysis of four English abstract nouns and their Spanish equivalents

This paper presents an analysis of the expression of evidentiality with the English nouns evidence, indication, proof and sign and their Spanish equivalents evidencia, indicación, prueba and señal. The nouns are described as shell nouns having the properties of encapsulating, signalling and labelling. The delimitation of their evidential and nonevidential uses is determined by three factors: existence of a qualified proposition (Belief), non-occurrence within an irrealis context and constant value of the evidential qualification when the Belief refers to a plurality of events. The difficulties posed by the delimitation illustrate the problems involved in determining the scope of evidentiality when expressed by lexical devices belonging to the content of a proposition. A quantitative analysis was carried out on 400 occurrences of the nouns, extracted from two comparable corpora. The results reveal that all the nouns except two expressed evidentiality in most cases, that the linguistic context in which they appear shows great variation in terms of syntax and information structure, and that the labelling function is prominent. The results also uncover idiosyncratic evidential expressions with some of the nouns.


Introduction
The abundant literature on evidentiality that surged after pioneering references such as Chafe and Nichols (1986) or Willett (1988) has mostly concentrated on grammatical markers of evidentiality (Aikhenvald 2004, among many others) and certain types of words or expressions such as adverbs and lexical verbs of perception or cognition (Celle 2009;Hennemann 2012;Ruskan 2015;Usonienė & Šinkūnienė 2013;Wiemer & Socka forthcoming).By contrast, this paper explores evidential expressions containing nouns, which have received comparably scant attention.The contribution of this research to the literature on evidentiality is twofold.Firstly, it uncovers ways of expressing evidentiality which, in functional approaches to this category, might well be added to those most frequently studied to date.Secondly, a number of factors are proposed for distinguishing evidential and non-evidential uses of the nouns under study; these factors seem to be useful for the delimitation of the scope of evidentiality when expressed by different kinds of linguistic devices belonging to the content of a proposition.
The nouns selected for the study have an evidential meaning of "a thing or set of things helpful in forming a conclusion or judgment". 1The English nouns chosen are evidence, indication, proof and sign, the four nouns considered in Schmid's (2000) monograph on shell nouns as "[t]he four workhorses in the 'Evidence' family", due to their high frequency in comparison with other similar nouns such as signal, implication or symptom in his corpus, the British section of Birmingham's Bank of English.2For Spanish, the four respective correlates evidencia, indicio, prueba and señal were selected.It might be argued that the nouns indicación and signo are also correlates of indication and sign, respectively, but their meaning is more general and they seem to display a high number of non-evidential cases: indicación often refers to an order to do something, and signo to a physical symbol, as in the case of arithmetic and punctuation signs; señal and especially indicio were considered to have comparably more evidential uses.The eight nouns were studied qualitatively for their discourse properties as shell nouns and the distinction between evidential and non-evidential uses, and were then analysed quantitatively according to these latter two uses, as well as to other syntactic and discourse properties of their evidential uses.The analysis was carried out on occurrences extracted from two comparable corpora, the British National Corpus for English and the Corpus de Referencia del Español Actual (CREA) for Spanish (see Section 4 below).Since the English and Spanish nouns have the same initials, the acronym 'EIPS' will be used to refer to the eight nouns.The expressions 'English EIPS' and 'Spanish EIPS' will refer to the nouns in the respective languages.
The structure of the paper is as follows: Section 2 specifies the concept and scope of evidentiality adopted in this paper, and the role of evidential nouns therein.Section 3 concerns the discourse properties of EIPS as a subtype of shell nouns.Section 4 describes the corpora and the data analysed.Section 5 deals with the distinction between evidential and non-evidential uses of EIPS.Sections 6 and 7 cover the quantitative analysis: Section 6 describes the database, and Section 7 specifies and discusses the results.Finally, Section 8 summarises the main conclusions and proposes suggestions for further research.
2 Evidentiality: concept, scope and role of evidential nouns Considering references such as Willett (1988), Nuyts (2001Nuyts ( , 2009)), Boye & Harder (2009), Wiemer & Stathi (2010), Boye (2012) and Carretero & Zamorano-Mansilla (2013), I propose a definition of evidentiality as the linguistic expression of the kind, source and/or evaluation of the evidence that someone, typically but not necessarily the speaker/writer, has or claims to have at his/her disposal, for or against the truth of the proposition. 3According to this definition, the bold expressions in examples (1-3) are evidential expressions;4 the qualified propositions are underlined: (1) 'Whereas Henrietta, I see, ' Ivan continued with a remorseless pity, 'is very much at home here.' (2) Shirley MacLaine, according to her autobiography, similarly relies on contacting disembodied entities through various mediums.
(3) The schools were evidently back after Easter.
In (1), I see indicates that the evidence in favour of the truth of the underlined proposition is of a perceptual kind (more specifically, visual).In (2), the expression in bold indicates that the evidence for the truth of the proposition is of a communicative kind, and that its source is Shirley MacLaine's autobiography.The inclusion of 'evaluation' in the definition of evidentiality is more controversial: some authors, such as De Haan (2005, 380) or Wiemer & Stathi (2010, 276), state that the expressions that evaluate the evidence for or against the truth of the proposition should be considered as epistemic.I would say, however, that evaluation of the evidence should be distinguished from evaluation of the truth of the proposition itself: evidently in (3) indicates evaluation of the evidence as strong, resulting in high commitment to the truth of the proposition; by contrast, 3 Throughout the paper, the term 'truth' is not used as a theoretical notion (as in truthconditional semantics), but as a pre-theoretical concept that people use in order to come to terms with the world: for example, in many contexts the utterance 'This pullover is red' would be considered as true even if the speaker/writer knows that it is not an absolute truth, since colour is an impression produced in the retina by light beams, not shared by certain animals or colour-blind people.
no doubt in its place would indicate high commitment to the truth of the proposition without allusion to evidence, thus being 'epistemic' in the strict sense.Expressions of the 'evidently' type might well be considered as 'epistentials', a term increasingly used in the literature (Simon-Vandenbergen & Aijmer 2007;Usonienė & Šinkūnienė 2013;Lampert 2015).
It must be noted that, for an expression to qualify as evidential, its syntactic scope need not be clausal: as Boye (2012, 183-184) points out, a non-clausal expression may communicate an implicit proposition, which can be made explicit by means of a paraphrase with clausal scope.For example, in (4) the syntactic scope of apparently is the adjective insoluble, but the implicit proposition qualified is 'the problem is insoluble' and the adverb is therefore evidential: (4) The government is faced with an apparently insoluble problem.
Another feature of the approach to evidentiality used here is its status of functionalconceptual domain "without interference from structural criteria associated with different forms of coding" (Boye & Harder 2009, 10).It is worth remembering that the treatment of evidentiality in early references, such as works by Franz Boas and Roman Jakobson in the first decades of the 20 th century, and also in most of the papers in the seminal book edited by Chafe & Nichols (1986), was restricted to grammatical realisations.In a similar fashion, Anderson (1986,(274)(275) considers evidentials as a "special grammatical phenomenon", and proposes four criteria that evidentials need to meet to be considered as such: the first states that "[e]videntials show the kind of justification for a factual claim which is available to the person making that claim <…>"; the second, that evidentials cannot be the main predication of the clause; the third, that evidentials must indicate evidence as their primary meaning, not only as a pragmatic inference; and the fourth, that evidentials should be "inflections, clitics, or other free syntactic elements (not compounds or derivational forms)".Studies on evidentiality in this narrow sense have facilitated the understanding of its status as a semantic category and its different ways of expression in a wide range of languages in the world, and have shed light on structural facts about these languages.Here I adopt a complementary approach, including all the devices that mean evidentiality in the sense of the definition stated at the beginning of this section, in order to cast light on its scope as a functional domain.In accordance with this perspective, the evidential devices may be grammatical, lexical, semantic or pragmatic: only the first of Anderson's criteria needs to be met for an expression to qualify as an evidential, with the caveat that 'claim' is to be understood as a qualification for the truth of a proposition, not as a representative speech act that consists in asserting that something is true in spite of the fact that people might not believe it. 5 As shown in Boye (2010Boye ( , 2012)), evidential qualifications fall under the scope of speech acts, not the other way round: for example, patently in ( 5) is an evidential qualifying the proposition 'the president is guilty', and this qualification lies within the scope of a rogative speech act (question): (5) Television screens around the world seemed to show that the president and his associates were guilty as charged and there could be no doubt that public confidence in the Nixon administration had been shattered.Why, therefore, did congress not get on with the business of removing a patently guilty president?
Many studies on evidentiality in Western languages (see the first paragraph of the Introduction) have concentrated on expressions that meet Anderson's criteria to some extent, such as adverbs and lexical verbs.These expressions fulfil Anderson's second criterion, since they are not the main predication of the qualified clause, and often the third criterion, even if some of them communicate evidentiality not as a semantic meaning but as a Generalised Conversational Implicature, such being the case of seem, which is not evidential when referring to misleading appearances as in 'John seems stronger than he really is'.EIPS are more peripheral devices for the expression of evidentiality: to start with, they are not evidentials per se, but may be part of evidential expressions or not, depending on the linguistic context in which they occur.For example, evidence in ( 6) is part of an evidential expression (the non-underlined part of the sentence), whereby the writer expresses a qualification of the truth of the proposition expressed by the underlined clause, based on the kind of evidence cited and with the strength conveyed by suggests; the degree of commitment is comparable to that of 'it seems that', the difference being that the Noun Phrase (NP) headed by evidence characterises the evidence with a much higher degree of precision.However, evidence is not evidential in any of its occurrences in (7), since it does not qualify any proposition: the fragment is an account of the possible ways in which evidence (treated as a commodity) can be taken in legal proceedings involving foreign parties.
(6) Seismic evidence of the internal structure of the delta suggests that over thirty events can be identified across the region.
(7) Evidence can be freely taken by agents acting on behalf of foreign litigants; but no compulsory processes may be used, nor may the evidence be taken on oath.A foreign court is at liberty to appoint a consul in England of its own country, or any other 5 This characterisation of the claim as a speech act is based on the entry for the noun 'claim' in the Cambridge Dictionary for Spanish Learners of English, accessed September 21, 2016: http://dictionary.cambridge.org/es/diccionario/ingles/claimperson it desires as an examiner to take evidence.So long as the witnesses are willing to attend to give evidence the examination may be completed and the result returned to the foreign court without the intervention of the court in England.
The evidential expressions containing EIPS may be considered as peripheral, since they fail to comply with Anderson's (1986) second and fourth criteria: they can be the main predication of the clause, as in (6) above, and do not belong to the categories specified in the fourth criterion.By contrast, they can be considered as complying with the third criterion for the reasons given at the end of Section 5.
The choice of EIPS as the subject of study in this paper agrees with the approach to evidentiality proposed in this section and aims to shed light on the different resources that English and Spanish have for the expression of evidentiality as a functional category.

The evidential frame
EIPS belong to the category of shell nouns (Schmid, 2000;Hunston & Francis, 2000), since they can all occur in the two criterial syntactic patterns indicated in Schmid (2000, 3): 1. Determiner + (Premodifier) + Noun + postnominal that-clause, wh-clause or to-infinitive; 2. Determiner + (Premodifier) + Noun + be + complementing that-clause, wh-clause or to-infinitive.Schmid (2000) sets forth a classification of shell nouns that consists of two levels.The first divides the nouns depending on the type of experience described.Accordingly, five groups are distinguished: -The factual group, which describes facts and states of affairs; -The mental group, which describes ideas, cognitive states and processes; -The linguistic group, which describes utterances, linguistic acts and their products; -The modal group, which describes possibilities, abilities, permissions, obligations, etc; -The eventive group, which distinguishes activities, processes and states.
The second level divides each group into subgroups in terms of uses rather than meanings.In this way, certain nouns that have developed different uses, such as point or position, can be characterised as belonging to more than one subgroup.Schmid includes evidential nouns in the factual group, arguing that they construe experience as a fact, even though he acknowledges that they share semantic features with the mental group.I believe that evidential nouns should be classified instead in the mental group or in the modal group: evidential nouns such as EIPS are similar to mental nouns such as belief, hope or fear and to modal nouns such as possibility and certainty, and different from other factual nouns such as fact, problem and reason, in that they are non-factive in the sense of Kiparsky & Kiparsky (1970).The contrast may be seen in the construction 'the + noun + that': the noun problem is factive since it triggers the presupposition that the proposition expressed by a following that-clause is true (8), while the evidential noun indication does not trigger the corresponding presupposition ( 9): (8) Moreover, it must be said that even the patristic position in no way solves the problem that the symbolism of Christ is somehow male.
(9) At 3pm, he will meet Clinton and the two will be together for the next six hours, an indication that the president regards the Prime Minister's visit as more than just a courtesy call.
Each of the subgroups is characterised by a different frame, a notion that Schmid has adopted from Talmy (1996) and may be defined as "a set of conceptual elements and relations that <…> are evoked together or co-evoke each other" (Talmy 1996, 238).The frame of evidential nouns consists of two components, the Sign (an observed fact) and the Belief (a mental state), and a relation that links them: the observation of the Sign triggers the Belief.For example, in (10), a paragraph about a bird, (10) A few minutes later it emerged and flew off and then when it returned it looked at the wall and saw the marks and went inside.Ample proof indeed that it had recognized them and it realized that it was the entrance to its home.
the Sign is "[the bird] saw the marks and went inside" and the triggered Belief is "that it had recognised them and it realised that it was the entrance to its home".In this paper, the evidential frame will be considered as such from a semantic point of view, without restrictions to given syntactic constructions: as will be seen below, evidential uses of EIPS display great variety in the syntax of both the Sign and the Belief.
In order to determine the status of the elements of a frame, Schmid (2000) adopts Lyons's (1977) distinction between first-order, second-order and third-order entities.First-order entities are physically perceivable things, such as persons, animals, plants or inanimate concrete entities; they have a concrete location in space and fairly constant perceptual properties.The distinction between second-order and third-order entities, usually called 'states of affairs' and 'propositions', respectively, is more difficult to grasp.Boye (2012, 278) states that the difference between both lies in that "states of affairs can be said to occur, whereas propositions can be said to have a truth value" (see also Wiemer & Stathi 2010).Some shell nouns can take states of affairs or propositions as complements.For example, the complement of suggestion in (11) denotes a state of affairs, which concerns the bringing about of an action and has no truth value, while the complement in (12) denotes a proposition, i.e. a piece of information of the world that may be true or false: (11) The next morning, a Sunday, he agreed to Dick's suggestion to see a West German doctor, had tests done the following Wednesday and was referred to Zurich for an operation a week later.
(12) In Jeremiah xliv.17 there is a suggestion that the moon was worshipped by the Israelites.
In the evidential frame, both the Sign and the Belief normally have the status of a proposition.They may be realised by clauses, as in ( 10), but this is not necessarily the case.They may be a plurality of propositions: in (10), both the Sign and the Belief consist of two propositions, separated by the conjunction and in both cases.As was seen in Section 2, the propositional scope may also be implicit.For example, the Belief may be realised by an NP and still have the status of a proposition, as in ( 13), (13) Hyperactive children may 'grow out of it' in time, but this takes a long time and their behaviour tends to get worse before it gets better.Their inability to concentrate or order their thoughts means that they generally do not learn much at school, even though they may be quite intelligent.Some have difficulty in writing and spelling.
There is evidence of criminality and psychotic behaviour in some hyper-kinetics when they reach adulthood, so it is advisable to try to sort out the problem sooner rather than later.
where the Belief, expressed by the underlined stretch, is paraphraseable by the clause "that some hyper-kinetics develop criminality and psychotic behaviour when they reach adulthood", which makes the propositional status clear.
Possible exceptions where the Sign is not a proposition are cases such as ( 14), an example cited from Schmid (2000, 111) which he considers to be no exception, arguing that "Gary Clark", is at first sight a first-order entity (a male person), but the Sign is really the fact that he "exists and behaves in a certain way": (14) Like the Postcard posse and Lloyd cole before him, Gary Clark is proof that the Great British Songwriting Tradition is alive and particularly well North Of the Border.
These uses of first-order entities as the Sign are clear cases of metonymy.In ( 14), the whole person 'Gary Clark' stands for some facts connected with his life, such as 'Gary Clark is Scottish' or 'Gary Clark wrote high quality songs' but not others such as, say, 'Gary Clark is bald' or 'Gary Clark has blue eyes'.As for the Belief, it is rarely realised by a first-order entity, an exception being the two cases found of the Spanish construction 'poner en evidencia a' (literally 'put in evidence') with an NP referring to a person.This construction has a pejorative meaning, with the sense of 'make it evident that someone's value or skill to do something is limited or unsatisfactory', as in ( 15 where the Belief, me, also has a metonymic value, standing for 'that I did not know those programming languages even by name', 'that my computer skills are weak, etc. 6As a last observation about the evidential frame, there is a difference between the constructions with EIPS in the two criterial patterns specified at the beginning of this subsection: the that-clause expresses the Belief in the construction 'Noun + that-clause' (16), and the Sign in the construction with 'Noun + BE + that-clause ', as in (17), where the Belief appears before the clause in which prueba occurs: (16) There is convincing evidence that a predisposition to alcoholism itself has a significant genetic component choose.ptcp.f.sg by unanimity.sg'I believe that this conference was not won by any sector over another.The proof is that the new executive committee was unanimously chosen.'

A semantic and discourse characterisation of EIPS as shell nouns
Schmid (2000) signals three major functions common to all shell nouns: encapsulating, signalling and labelling.The difference between the first and the second function lies in that 'encapsulating' takes an inward perspective and 'signalling' takes an outward perspective.In other words, the encapsulating function concerns EIPS' containment of longer chunks of information, thus "act[ing] as host and shelter for things that would otherwise easily be dispersed or damaged" (Schmid 2000, 13); 7 the 'shell' metaphor is most prominent in this function.Encapsulation gives rise to hypostatisation and reification, since the encapsulated information is conferred the status of an entity (see Conte 1996;Schmid 1997;Schmid 2000, 363-369).Schmid states that the nature of the resulting entities is illusory: in spite of the inherent stability provoked by conceptualisation 7 As Conte (1996) and Schmid (1997) have pointed out, the relation between the encapsulator and the encapsulated information is not always straightforward, since the encapsulated information is not always clearly delimited.I will not go more deeply into this issue here for reasons of space.
as an NP, the coded concepts are temporary and ephemeral.I agree that the entity so created is not as cognitively salient as a concrete entity such as a dog, a table or a cake; however, temporariness varies with the context: for instance, investigation in science or journalism involves the consideration of evidence as a solid entity.In the case of EIPS, the encapsulated information is the Sign.For example, evidence in (18) stands for the Sign expressed by the stretch from 'Dr Estelle Ramey' until the end of the paragraph.
(18) Stress, as everyone knows, is a real passion-killer in both women and men.As one friend complained: 'One doesnʼt just suffer from oneʼs own stress -as soon as our manʼs job is going badly weʼre the ones who get the backlash and that always seems to include no sex'.Now evidence shows that it's more than the fact that he has something on his mind to distract him from sex -itʼs an actual physical phenomenon.Dr Estelle Ramey, professor emeritus of physiology at Georgetown University School of Medicine, USA, explains that 'your system is so delicately balanced that itʼs very difficult for your body to make two types of hormones at once.So decisions are made in a bio-chemical fashion as to which is more immediately necessary.When the brain starts signalling production of stress hormones it begins to inhibit sex hormone secretions.' The signalling function concerns the role of shell nouns in information structure as indicators that have to be processed anaphorically or cataphorically as substitutes for the information that they contain.The encapsulating and signalling functions are jointly captured in Conte's (1996, 1) definition of encapsulation as a cohesive device whereby an NP functions as a paraphrase for a portion of a text.Conte's paper is restricted to anaphoric encapsulation, where the NP stands for a preceding stretch of discourse, and she states accordingly that the paraphrase is resumptive; on the other hand, in cataphoric encapsulation the NP stands for a following stretch of discourse, the paraphrase being therefore prospective, as in (18) above.Due to their encapsulating and signalling functions, shell nouns provide an economical means for processing large chunks of information and thus offer advantages for discourse organisation in comparison to the stretches of discourse that they replace.Schmid (2000, 377) even suggests that shell nouns reduce the amount of information held in an active state by discourse participants, thereby relieving the load on the short-term memory buffer.
The labelling function assigns the lexical meanings of shell nouns to the encapsulated information (cf.Conte 1996, 6).Thus shell nouns can manipulate the discourse, since the encapsulated information is given the status of an entity characterised by their meaning.
In the case of EIPS and the other evidential shell nouns, the encapsulated information (the Sign) is labelled as evidence for (or against) the truth of the Belief.
Schmid (2000, Chapters 14-17) deals with variations of the relative weight of each of the three functions depending on the construction where shell nouns appear.For reasons of space, I will not give an extensive outline of the discussion.Suffice it to mention two observations relevant to our purposes.The first is that the occurrence of the nouns in definite NPs as anaphoric encapsulators is mainly motivated by the linking function, i.e. by the need to give coherence to the discourse, as in ( 19), where the anaphoric encapsulation of the Sign (the information given in the stretch from 'the fact that…' until 'the headmaster's window') by the NP 'this evidence' facilitates information processing as well as information structuring in subsequent discourse: (19) Their assumption that 'he knows what he has done wrong' is based on the fact that Rover is slinking along the floor with its head and tail down looking for all the world like a naughty school boy who's just broken the headmaster's window.On the basis of this evidence they proceed to chastise Rover to a degree consistent with their temperament.
The second observation is that anaphoric encapsulation in the construction 'Sign + be + shell noun (phrase) (th-be-N)' places the noun after the copula, which is a communicatively prominent position, with the consequence that its meaning, and therefore the labelling function, is most prominent, as in ( 20):  As was stated in Section 2, EIPS have evidential and non-evidential uses.To start with, some of the nouns are polysemous and have non-evidential meanings.This is the case of sign and its Spanish correlate señal in the meaning defined by The Free Dictionary as "a posted notice bearing a designation, direction, or command: an EXIT sign above a door; a traffic sign" (original italics).8Several cases were found where señal referred to an instruction to do something; these cases belong to Schmid's (2000, 170-173) frame of 'order' within the linguistic group and are therefore non-evidential.In its turn, prueba also has the non-evidential meaning 'test, trial, exam'.
The evidential uses are those included in the definition of evidentiality provided in Section 2. That is, EIPS are part of evidential expressions when they assess the Sign as evidence for or against the truth of the Belief.This is the case of ( 21), where the NP headed by proof assesses the Sign (in round font) as evidence for the Belief (the proposition expressed by the underlined stretch): (21) passenger transport executives provided public proof of their interest (and subsidy) of metropolitan rail services by applying their own liveries.
Two objections may be made about the consideration of these cases of EIPS as realisations of evidentiality, but both will be rejected.The first is that EIPS always belong to the propositional content of the clauses in which they occur.In ( 21), the truth conditions of the whole sentence are different from the truth conditions of the Belief ("that they were interested (and were subsidising) metropolitan rail services").However, this fact does not invalidate their evidential status.As stated in Boye (2012, 197-198), a necessary condition for an expression to be evidential is to have a propositional scope, but this does not mean that the evidential itself, to be considered as such, cannot be part of a proposition.He illustrates his point with the example "It is evident that Bob is ready", where evident belongs to the proposition expressed by the whole sentence, different from the proposition under its scope ("Bob is ready").However, it must be noted, as will be seen below in this section, that this fact is a motivating reason why EIPS are not evidential per se, but indeed may be part of evidential expressions, in contrast to expressions such as evidently or allegedly, which need not be part of any proposition and are therefore evidential in most cases.
The second (and related) objection, connected with Schmid's (2000) inclusion of evidential nouns within the factual group, is that EIPS code the relation between the Sign and the Belief as a fact of the world.However, this factual status does not prevent the relation from expressing an evidential qualification of the truth of the proposition; rather, it is a necessary condition for evidential uses of EIPS.9For example, proof in (21) above is encoded as a fact and expresses the evidential qualification that the Sign is a proof for the truth of the Belief.
This demonstration of the existence of evidential cases of EIPS is to be followed by a delimitation of criteria for distinguishing evidential from non-evidential cases.The examination of the occurrences chosen for the quantitative analysis has uncovered two essential criteria for considering each of them as evidential: 1) the Belief exists, and 2) the evidential qualification is constant.These criteria will be discussed in the remainder of this section.
The criterion of existence of the Belief supports the consideration as non-evidential for the cases of evidence in (7), since the evidence is treated as a commodity not geared to any particular Belief.By contrast, the Belief ultimately exists in the following types of cases, which have been consequently considered as evidential: A) When the Sign is not explicit.For example, in ( 22) the existential construction evaluates the truth of the Belief by indicating that it is supported by evidence, without further specification.
(22) Timing control.This allows the designer to plan the way in which the user will interact with the program in real time.There is evidence of the importance of timing in stimulating pupil learning.The micro will allow this both to be used and further explored.
B) When the Sign is non-existent.In these cases, EIPS communicate an evidential qualification by evaluating the evidence for the truth of the Belief as non-existent (23): (23) Now it is not so easy to voice those arguments.When John Major complained of the relentless diet of violence on our screens he tapped into a real public concern.That is why Alan Yentob, controller of BBC1, felt duty bound to reply, publicly apologising for an episode of Casualty which featured a riot.But Yentob pursued an ethical balancing act, still arguing there was no proof of a causal link.This father of a three-year-old also said parents should control what their children watched -as he did.
C) When the evidential relation between the Sign and the Belief is not clearly signalled by syntax.The relation between the two components of the evidential frame is to be understood in a broad semantic sense, not in a narrow syntactic or semantic sense.For example, the that-clause in ( 24) is an apposition of presumption and not of evidence.However, the proposition expressed by this clause, together with the initial clause, semantically counts as the Belief of evidence, considering that the expression 'to support the presumption' has been introduced basically to lay emphasis on the non-factual status of the Belief.Therefore, (24) has been counted as an evidential case of evidence.
(24) The attitudes set in late Victorian times can be traced in British industry right through into our period.Perhaps most striking, has been their persistence even in the new science-based industries of the twentieth century <…>.It is not possible, of course, to be certain of the extent of their deleterious effect without a great deal more investigation.But there is already much evidence to support the presumption that the effect was pervasive.The contrasts which can be drawn with foreign industry even with such a tradition-bound country as France -are often stark.The widening of educational opportunity in the immediate post-war period was not accompanied by radical changes in its content.
D) When the Belief is completely or almost completely implicit.In the occurrence of proof signalled in bold in ( 25), the speaker (PS28R) does not specify the Belief; however, its retrieval by the addressee (PS28P) from the situational context is made obvious in the next turn, where s/he uses the same noun proof and makes the Belief explicit: ( E) When the evidential qualification is not presented as stemming from the speaker/ writer at the moment of performing the utterance.For example, proof in ( 26) is within the scope of indirect reported speech.These cases could be considered as 'descriptive' in the sense of Nuyts (2001Nuyts ( , 2009)).However, the distinction between these 'descriptive' cases and the others (which Nuyts names 'performative') is not always clear: in ( 26), the word proof may have been introduced by the reporter or it may lead to interpret that the reporter shares the respondents' consideration of the exercise as proof for the truth of the proposition expressed by the underlined stretch.This difficulty to delimit both kinds of cases is also illustrated by ( 27), where the speaker uses proof as part of an evidential qualification given in the past but valid at the speech moment.
(26) Several respondents commented that both foster parents and children had enjoyed participating in an exercise which provided them with concrete proof of achievements and which sought to monitor progress systematically.
(27) 'You are not, as you believe, the offspring of Will Halidon-but of my late husband John Mowbray, fourth Duke of Norfolk.' Joanʼs face paled and she gazed wide-eyed at the duchess, as if suspecting that lady had taken leave of her senses.'Nay, it cannot be,' she breathed.'Bess Halidon, your mother, gave me proof of it when she came here on the day prior to her death and asked that you be found employment.
By contrast, the criterion of existence of the Belief rules out cases in which EIPS lie within the scope of irrealis, and are therefore not facts like in other cases.This is the case of ( 28), where indication concerns a Sign that was expected but was not finally produced, and consequently the proposition expressed by the underlined stretch is not a Belief triggered from it: (28) He chose as the topic for his 30 minute address 'The Revival of British Manufacturing Industry'.From the title, some might have expected Lilley to give an indication of his hopes for revival in the near future.But instead he spoke almost solely about the revival in the Eighties. 10  The other criterion, constant evaluation, applies to cases where the Belief contains reference to a plurality of entities, so that the Belief is a proposition with an open variable.For example, in ( 29), ( 29) I take the point Chairman, and, and suggest by the criticism we will try and get more explicit detail in the future.<…> In salaries or joint finance funding, so where there are differences it is almost entirely down to those two factors.It does give members an indication of the level of expenditure on particular services.
the NP particular services causes the Belief to be an open variable, paraphraseable as 'of which the level of expenditure on particular services is' or, alternatively, 'that for service X the level of expenditure is XX', 'for service Y the level of expenditure is YY', and so on.The beginning of the NP 'an indication', together with the preceding part of the sentence (which has emphatic affirmation) qualifies the Sign ('it', which probably refers to a report carried out by the speaker and others) as valid evidence, 10 This exclusion is associated with the tendency for evidentials to occur in realis clauses rather than irrealis clauses and presuppositions, attested by Anderson (1986, 277).
without differentiating between the reliability of indications for each particular service, so that this use of indication is considered as evidential.
In other cases, however, EIPS do not provide a constant qualification for all the propositions included in the Belief, as in ( 30), where 'the indirect evidence' does not provide a constant qualification for all the paintings: the indirect evidence of critical approval may be strong for painting A, weak for painting B, intermediate for painting C, and so on.Consequently, evidence is not evidential: (30) Long catalogue entries recall the cynical remark that the price of a painting is in ratio to the length of the bibliography in the sale catalogue.
Besides seeking to establish the authenticity of works in the sale room, auctioneers are also concerned that the title to ownership is secure.The indirect evidence of critical approval by a succession of owners may be interesting to a reader, though admittedly this demands some special knowledge about the standing of the collectors concerned.
The previous paragraphs of this section lead to infer that the distinction between evidential and non-evidential uses of EIPS is complex, which is in all probability due to their peripheral status within the domain of evidentiality, in the sense that the nouns always belong to the contents of a proposition and lie outside Anderson's (1986) second and fourth criteria for being evidential.It may well be argued that evidential uses of EIPS fulfil Anderson's third criterion: the evidential qualification is triggered by the meanings of the individual words and their combination in the linguistic context, not by conversational implicature, and it is non-defeasible, thereby belonging to semantics, not to pragmatics.

The database
The 400 occurrences of EIPS selected for the quantitative analysis were registered in a database, created with the Excel program to find the number of cases in which each noun is part of an evidential expression, as well as the frequency of different syntactic configurations of the evidential frame, common collocations, and the role of the nouns in information structure.The fields of the database are as follows: a) The actual example.
b) The evidential or non-evidential status of the example, according to the criteria in Section 5.The remaining fields were analysed only for those cases with evidential status.c) Clue.This field registers whether there is a clue that nuances the value of the evidence by strengthening or weakening it.The clue consists mostly of constituents of the NP headed by EIPS, such as adjectives (clear, plentiful, sure, untrustworthy…) and determiners (some, little, no…).Other expressions have also been found, such as the adverb indeed in (31): (31) But even in the 1950s and '60s the unemployment figures may not have reflected the 'true' number unemployed because of people's failure to register.In particular this applied to women, and one indication of it is indeed the change over the post-war period in economic activity rates among women.d) Sign.This field indicates whether the Sign is explicit or not.In the former case, its syntactic role is registered if it is in the same clause as the evidential noun; otherwise, the field only registers whether the Sign occurs before or after the clause of the evidential noun.e) Belief.This field indicates the syntactic function of the Belief when it is in the same clause as the evidential noun.When the Belief is realised in a different clause, this field registers whether it occurs before or after the clause of the shell noun.In the case of postmodifiers of the NPs headed by EIPS, a categorial distinction was made between Prepositional Phrases and clauses, taking into account a syntactic difference between the two languages: in English, appositive clauses are realised by that-clauses after the noun, as in 'the proof that David is innocent…'; in Spanish, however, appositive clauses have to be preceded by the preposition de, as in 'la prueba de que David es inocente', literally 'the proof of that David is innocent'.To make the results of the two languages comparable, the patterns mentioned above were both counted as appositive clauses, and the syntactic difference brought about by the Spanish preposition was not considered.
f) NP of the shell noun.This field registers the syntactic function of the NP containing EIPS within the immediate superior hierarchical unit.The following kinds of clausal constituents have been distinguished: Subject; Notional Subject (in existential constructions, as in 'There is evidence that…'); Direct Object; Subject Complement ('X is evidence that…'); Object Complement ('I consider X as evidence that…'); Prepositional Complement, i.e. non-attributive complements governed by the verb and headed by a preposition ('X relied on the evidence…'); Adjunct (optional clausal constituent); when the NP is not a clausal constituent, it may be a postmodifier of another constituent ('conscious of the evidence that…') or an apposition after the clause ('I saw X, an indication of…').
g) Thematicity.This field specifies whether the NP containing EIPS belongs to the Theme or the Rheme of the clause in which it occurs.Theme, the point of departure of the clause, is realised by first position, and Rheme is the rest of the clause.The approach followed is that of Lavid et al. (2010), according to which the Theme can comprise either one syntactic constituent, for example a Subject (labelled Thematic Head), or more than one element, as when a clause starts with an Adjunct (Pre-Head) followed by the Subject (Thematic Head). 11The rest of the constituents make up the Rheme.The field registers whether the NP containing EIPS is (part of the) Pre-Head, Thematic Head or Rheme, no matter whether it is an immediate clausal constituent or not.The appositive cases are considered part of the Rheme, since they appear at the end of sentences.In the literature on information structure, there is almost unanimous agreement (Firbas 1992;Halliday & Matthiessen 2014;Lavid et al. 2010) that in English and Spanish the more important information tends to occur in late parts of the clause, so by and large, the Rheme may be considered as having more communicative weight than the Theme.

Overall findings: results and discussion
To start with, the data selected for both languages display more instances of evidential than of non-evidential uses, the crosslinguistic difference being small (see Table 1).
The distribution was more uniform for the English nouns: for all of them the evidential examples totalled from 30 to 36 (between 60 and 72% of the cases), while the Spanish data showed a higher frequency for evidencia and indicio and a lower frequency for prueba and señal, the only nouns that display more non-evidential than evidential occurrences.Within the evidential occurrences of EIPS, the cases with a clue nuancing the evidential meaning were also counted, and the percentages calculated in relation to the total number of evidential cases.The clue is common in both languages, which indicates that language users are careful to nuance the degree to which the evidence supports the truth of the proposition.Clues are clearly more frequent in the English data.The most often nuanced nouns are evidence and sign, while the least often nuanced noun is by far señal, followed by prueba.
The realisations of the Sign in the English and Spanish EIPS are specified in Tables 2 and 3. Their distribution is quite similar in the data of the two languages, although a number of differences are easily noticed.Firstly, the English determiner no is quite frequent, while its Spanish equivalent ningún (counting its inflections for feminine and plural) is much rarer, and limited to indicio: lack of evidence is expressed in other ways, such as negation of the main clause.Secondly, the Spanish data display more cases of the Sign fulfilling the functions of Subject and Adjunct in the clause where the evidential noun appears; and thirdly, the English data show more cases with the Sign in a clause following the clause of the evidential noun.Some peculiarities of the Sign occurring with individual nouns are the high frequency of the determiner no with sign, and also the relatively high frequency of evidencia and indicio as heads of NP Subjects: 10 of the 14 cases of Signs-Subjects with evidencia occurred in the construction poner/dejar en evidencia ('put/leave in evidence'), where the Belief is expressed by the Direct Object.No cases were found of the English equivalent construction put in evidence, although it is grammatically correct.With indicio, all the cases of Signs-Subjects occur in a copular construction where the noun is the head of the Subject Complement and is postmodified by a Prepositional Phrase with de that introduces the Belief, as in ( 33 The Belief also displays a wide variety of realisations in both languages, as illustrated in Tables 4 and 5 The realisations of the Belief in a clause preceding the one containing the shell noun occurred above all with evidence and proof in the English data, and with evidencia in the Spanish data.Occurrences of the Belief in a clause following the clause where the evidential noun occurred were only found in the Spanish data, and determiners were only found with evidencia ('this/that evidence' with the sense of 'evidence of this/ that').The cases of Direct Object are much more common in the Spanish data, due to the occurrences of the construction poner/dejar en evidencia ('put/leave in evidence') mentioned above.
Tables 6 and 7 8 and 9 specify the cases in which the NPs headed by EIPS are (part of) the PreHead, the Head or the Rheme of their clauses.The results indicate that the cases in which NPs are part of the Rheme heavily outweigh those in which they are PreHeads or Heads, so EIPS may be considered prone to having communicative weight.Additionally, even when EIPS are PreHeads or Thematic Heads, the high frequency of postmodifiers confers informational weight to them, since the clausal constituent in which they occur is long.In sum, although a more in-depth analysis of the data would be required to confirm this statement, EIPS are not easily seen as transmitting only 'Given' information, i.e. information treated as known by or accessible to the addressee.EIPS do not seem to share a property that Schmid (2000, 308) attributes to shell nouns in general: "More often than not, the characterisation entailed in a given noun goes entirely unnoticed, since one tends to overlook that the choice of shell nouns is completely up to the speaker of an utterance."Rather, EIPS seem to have a prominent labelling function, whereby language users clearly assign the status of evidence to the encapsulated information.This paper has explored the expression of evidentiality, understood as a functionalconceptual domain characterised by the meaning of kind, source and/or evaluation for or against the truth of the proposition, with the English nouns evidence, indication, proof and sign and their Spanish equivalents evidencia, indicio, prueba and señal.The eight nouns are shell nouns, and as such they have the discourse functions of encapsulating, labelling and signalling.More concretely, they belong to Schmid's (2000) subtype of evidential nouns, characterised by an evidential frame consisting of an observed fact (the Sign), a mental state (the Belief) and a relation between both, namely that the observation of the Sign triggers the Belief.The nouns do not have evidential status per se; rather, they have evidential and non-evidential uses, depending on the linguistic context in which they occur.Evidential uses are characterised by the existence of the Belief, non-occurrence of the nouns in the scope of an irrealis context, and constant value of the evidential qualification when the Belief refers to a plurality of events.If the uses meet these conditions, the evidential meaning is non-defeasible and may therefore be considered as part of semantics, not of pragmatics.However, these uses are to be considered as peripheral expressions within the domain of evidentiality, since they are always part of the content of a proposition and do not fulfil Anderson's (1986) second and fourth criterion.However, they are definitely to be included in functional-conceptual accounts of evidentiality, since they are used by the speaker/writer to qualify the truth of a proposition by characterising and evaluating the evidence for or against this truth.
The quantitative analysis of 400 occurrences of the nouns extracted from the BNC and the Peninsular Spanish part of the CREA has uncovered that all the nouns except prueba and señal occurred in evidential uses in more than half the cases, that the evidential qualification is often nuanced with other expressions (clues), and that there is great variation in the Sign and the Belief, in terms of both syntax and information structure: both can be realised by different kinds of constituents in the same clause as the evidential noun, or else they can be located in previous or following clauses.The tendency of the nouns to occur in clausal Rhemes suggests that they have a weighty labelling function, in the sense that the status of the Sign as evidence for the truth of the Belief is highlighted.The study has also uncovered idiosyncratic uses of some of the nouns, as in the constructions en señal de 'as a sign of' and poner/dejar en evidencia 'put/leave in evidence'.
Similar research could be carried out on more evidential nouns (such as signal or symptom) and the number of languages could also be increased.Studies might also be carried out on other nouns that are not evidential themselves but are easily conceived as forming part of evidential strategies (for example, with the verbs show or suggest), such as nouns with a meaning of information aimed to be used as evidence (data, figure, statistics, etc.) and factual shell nouns of the 'linguistic' type (report, rumour, statement, etc.).The factors signalled here for distinguishing evidential and non-evidential uses of the nouns under study might also be tested for expressions of other syntactic types that also belong to the content of a proposition, such as adjectives.For example, the irrealis factor seems to block evidentiality in a number of adjectives such as apparent, as can be seen in ( 34): (34) Consequently, great care has to be taken when interpreting marks seen on aerial photographs, and wherever possible the site should be visited on the ground to establish whether the marks have resulted from a buried archaeological site or from some other cause.

( 20 )
And Graham acknowledged: 'He accepted what was said, and got on with the job.That is the sign of a good professional.'4The dataThe quantitative analysis was carried out on 50 authentic occurrences of each noun, totalling 400 examples.In all the cases selected the noun form was singular; plural forms (indications, proofs, etc.) were excluded.The 200 examples of English EIPS were retrieved from the British National Corpus (BNC), which contains approximately 100 million words, 90 per cent of which are written language and 10 per cent spoken language.Most of the texts were produced from 1975 onwards.The examples were randomised without a distinction between spoken and written language, by means of the option 'download random set' in the Sara-32 software, which selects a randomised set of examples of any size.The 200 examples of Spanish EIPS (again, 50 of each noun) were obtained from the Peninsular Spanish part of the Corpus de Referencia del Español Actual (CREA), which consists of texts produced between 1975 and 2004, totalling approximately 85 million words.This part of the CREA resembles the BNC in that the distribution is also 90 per cent written language and 10 per cent spoken language; therefore, the written and the spoken parts contain approximately 76.5 and 8.5 million words, respectively.In searches of words and expressions, the total number of cases is divided into pantallas ('screens') of 25 examples.Since the CREA does not have a randomisation facility such as Sara-32, randomisation was achieved by selecting a similar number of examples from each screen until a total of 50 was reached.A sample search is provided in Figure1, which shows that the total number of cases of indicio is 484, divided into 20 screens.I selected three examples of the screens with odd numbers and two of those with even numbers, thus totalling 50 examples.

Figure 1 .
Figure 1.Screenshot of the search of the cases of indicio in the Peninsular Spanish part of the CREA Do you know any programming language?¿C, ADA, Pascal?-he went on asking.-Well, the truth is that I have no idea of what you are talking about -I acknowledged, almost embarrassed that an eleven-year-old boy should put me in evidence.'

Table 1 .
Total of evidential and non-evidential cases of the EIPS analysed, and of evidential cases with a clue 11 Other possible constituents of the Theme are the Interpersonal Theme and the Textual Theme, but no examples were found of EIPS being part of them.

Table 2 .
Realisations of the Sign with the English EIPS

Table 3 .
Realisations of the Sign with the Spanish EIPS Some differences in individual nouns must also be noted.The high number of 'others' for señal is due to the construction en señal de or como señal de 'as a sign of', where the Sign comprises more than one constituent.For example, in (32) the Sign comprises the stretch from the beginning of the clause up to 'como', thus consisting of more than one constituent, and the Belief is expressed by the NP following of:

Table 4 .
. It occurs most frequently as a postmodifier with the preposition of Realisations of the Belief with the English EIPS in English and de in Spanish followed by an NP.This realisation is especially frequent with indicio and sign, followed by indication and señal, while it is less common with evidencia and prueba.The next kind of realisation, appositive clauses, is less than half as common in the two languages.

Table 5 .
Realisations of the Belief with the Spanish EIPS

Table 6 .
register the frequency of the different syntactic functions of the NPs headed by EIPS.The numbers lead to infer that the English EIPS, in contrast to the Spanish EIPS, commonly occurred with existential constructions, in the pattern 'there + verb + (determiner and/or adjective) + evidential noun'.The number of Direct Objects in the English data more than doubles that of the Spanish data; proof is frequent in this pattern (as in I have concrete proof, the snow gave them proof, we have no proof, or examples (21) and (27) above), while no cases were found of its correlate prueba.On the other hand, Adjuncts are more frequent with the Spanish EIPS, due to the construction en señal de discussed above.The function of Subject Complement is quite common in both languages, and some cases were also found of Object Complements.Prepositional Complements are much more common with Spanish EIPS, due to the construction poner/dejar en evidencia mentioned above.A few occurrences were also found in both languages where the NPs headed by EIPS function as appositions or as postmodifiers of other constituents.Syntactic functions of the NPs headed by the English EIPS

Table 7 .
Syntactic functions of the NPs headed by the Spanish EIPS With regard to thematicity, Tables

Table 8 .
English EIPS as (part of) the PreHead, Thematic Head or Rheme

Table 9 .
English EIPS as (part of) the PreHead, Thematic Head or Rheme