The continuum problem: Modified Occam's Razor and conventionalisation of meaning. %n order to overcome these problems % propose to reinterpret implicatures in terms of implicature'li(e effects delivered b! non' inferential processes#.
The continuum problem: Modified Occam's Razor and conventionalisation of meaning. %n order to overcome these problems % propose to reinterpret implicatures in terms of implicature'li(e effects delivered b! non' inferential processes#.
The continuum problem: Modified Occam's Razor and conventionalisation of meaning. %n order to overcome these problems % propose to reinterpret implicatures in terms of implicature'li(e effects delivered b! non' inferential processes#.
NOTICE: This is the authors version of a work that was
accepted for publication in <International Review of Pra!atics"# Chanes resultin
fro! the publishin process$ such as peer review$ editin$ corrections$ structural for!attin$ and other %ualit& control !echanis!s$ !a& not be reflected in this docu!ent# Chanes !a& have been !ade to this work since it was sub!itted for publication# ' definitive version was subse%uentl& published in INTERN'TION'( RE)IE* O+ PR',-'TIC.$ /012$ 3415$ 6OI:http:77booksand8ournals#brillonline#co!7content78ournals714#11/971:;;914<=44/41449 The continuum problem: Modified Occam's Razor and conventionalisation of meaning Abstract: According to Grice's Modified Occam's Razor in case of uncertaint! bet"een the implicature account and the pol!sem! account of "ord uses it is parsimonious to opt for the former# $o"ever it is "idel! agreed that uses can be partially conventionalised b! repetition# This fact % argue raises a serious problem for MOR as a methodological principle but also for the substantial notion of implicature in le&ical pragmatics# %n order to overcome these problems % propose to reinterpret implicatures in terms of implicature'li(e effects delivered b! non' inferential processes# )e!"ords: le&ical meaning* semantics+pragmatics* implicatures* associative processes 1. Introduction %n ,aul Grice's frame"or( for pragmatics the principle called Modified Occam's Razor -MOR. is an important means to cope "ith the demarcation bet"een semantic and pragmatic phenomena# Grice has emplo!ed it in order to argue for a pragmatic treatment of standard uses of certain linguistic forms: indefinite descriptions the counterparts of logical connectives in ordinar! language etc# The assumption / is that in the e&planation of such cases 0enses are not to be multiplied be!ond necessit! -Grice /123: //3'//1.* in other "ords one should not ma(e an appeal to a special sense for those e&pressions insofar as a general e&planation can be supplied based on the reasonable h!pothesis that the use at issue can be derived from a basic meaning through the application of some pragmatic principle# Recentl! the reliabilit! of MOR has been called into 4uestion "ith arguments that converge to"ards a common point that is the difficult! to dra" a straight line bet"een fi&ed le&ical meanings and pragmatic inferences -5ontl! 6778* ,hillips 67/6* Mazzone 67/9 a.: let us call it the continuum problem# :hatever the impact of these arguments on MOR it could be argued that research in pragmatics is not strongl! affected b! them# After all MOR is ;ust the application of a methodological principle of parsimon! to a specific domain and ever!one (no"s that methodological principles are appealed to onl! as a last resort "hen no direct evidence is available# / As a matter of fact scholars in pragmatics usuall! appeal to a variet! of more specific criteria and tests in order to assess "hether the use of a linguistic form is better e&plained as a pragmatic inference rather than as a le&icalised meaning# $o"ever there are reasons to thin( that the continuum problem has more serious and straightfor"ard conse4uences on pragmatic theories than simpl! ma(ing apparent the "ea(nesses of MOR# %n fact not onl! is the boundar! bet"een le&icon and pragmatics difficult to dra" on methodological grounds but there also seems to be as a matter of fact a continuum of phenomena "here Gricean pragmatics attempts to dra" that boundar!# Therefore % "ill claim the continuum problem does not onl! affect the reliabilit! of MOR as a methodological principle it also imposes empirical constraints on actual pragmatic theories < especiall! on theories concerning the relativel! recent research field of le&ical pragmatics# The e&istence of intermediate cases bet"een pure le&ical coding and full'fledged pragmatic inferences is recognised b! Grice himself as "ell as b! the research tradition stemming from his "or(# As a matter of fact MOR has been mostl! introduced in order to manage a class of cases that cannot be easil! e4uated to protot!pical -i#e# particularised conversational. implicatures since -and to the e&tent that. those cases involve standard uses of "ords regardless of specific conte&ts of utterance# On the other hand the intuition is that those uses are not part of the conventional meaning of "ords either# %n order to account for such cases Grice introduces the notion of generalised conversational implicatures -G=%s. that is genuine implicatures "hich nonetheless are normall! conve!ed b! specific forms of "ords unless / MOR is an instantiation of the more general Occam's razor "hich is used in science as an epistemological heuristic that is a provisional "a! to select bet"een competing h!potheses in case decisive evidence is lac(ing# 6 the conte&t suggests other"ise# To be more precise Grice's account is finer'grained than this# $e also envisages another (ind of implicature
conventional implicature
"hich "ould be a case in "hich part of the conventional meaning is nonetheless implicated rather than said#
This
notion
of
conventional
implicature
is not much credited in the current debate and % "ill leave it aside# $o"ever in the last decades man! scholars have accounted for the e&istence of intermediate cases bet"een purel! conventional and purel! inferential -implicated. meaning in terms of some sort of conventionalisation# This is in fact suggested b! 0earle -/128. "ho spea(s in this sense of conventions of usage that must be (ept distinct from meaning conventions < and therefore from le&ical meanings# %n the same vein Morgan -/123. has introduced the notion of short'circuited implicatures that is implicatures that are some"hat conventionalised as a result of past use -see also $orn and 5a!er /13>.# 5ach -/118 /113. has proposed that the term conventionalisation should be reserved for cases in "hich precedent serves to ma(e a given use possible "hile one should spea( of standardisation < and standardised implicatures < "ith regard to cases in "hich precedent merel! facilitates the inferential process licensing a certain use -5ach /118: ?22.# As ma! be seen in all these approaches the intuition of intermediate cases bet"een particularised implicatures and conventional meaning is accounted for in terms that are ver! close -as is the case for 5ach's standardisation. if not identical to the notion of conventionalisation# A ver! natural cognitive interpretation of this fact is based on the assumption that senses are conventionalised -i#e# coded in individual minds as conventional meanings. as a function of regularit! of use: ceteris paribus 6 the more a "ord form is used "ith a certain sense the stronger is the association bet"een the "ord form and the sense and the more "e perceive and treat the sense as a conventional meaning of that "ord form# As a conse4uence a "ord meaning can be partially conventionalised and this could e&plain the e&istence of intermediate cases bet"een particularised implicatures and full! conventional meaning# The assumption at issue has strong grounds in "hat "e (no" about associative learning and it could also have a role in the e&planation of ho" full! conventional meaning can come to be fi&ed at the level of individual cognition# %t is far from clear ho"ever that the pragmatic accounts "e considered above are compatible "ith such a reasonable assumption insofar as the! impl! a strong discontinuit! bet"een the recover! of le&ical meanings and the mechanisms delivering pragmatic inferences# %n fact "hile it seems plausible that "ord uses can be partiall! conventionalised it is less than clear that the notion of implicature admits 6 The importance of this clause "ill be clarified at the end of the fourth section# 9 of grades and therefore that it might complement conventionalisation as it is e&pected to do# %ntuitivel! if a certain "ord use is in the first place licensed b! a pragmatic inference and if this occurs again and again then that use should become progressivel! conventionalised# 5ut no one to m! (no"ledge has ever e&plained "hat "ould happen to inferential mechanisms "hen the implicated use becomes more and more conventionalised# @o the! decline to the ver! same e&tentA This does not seem 4uite plausible# And postulating categories of implicatures "hich are intermediate bet"een full! conventional coding and pure conte&tual implicatures does not solve the problem# Birst as "e "ill see such notions are debatable: the! aim to satisf! a theoretical need but "ith "ea( ;ustifications on cognitive grounds# And second the problem is ;ust postponed since one "ould as( again ho" the notion of standardised implicatures or the li(e might ta(e into account the fact that uses can be more and more conventionalised# Must "e presume that the need for pragmatic -though standardised. inferences declines proportionall! in a graded fashionA Or else is it the case that the more standardised implicatures become usual the less inferential < so to spea( < the! areA One point should be emphasized# % do not argue against the possibilit! to dra" conceptual distinctions such as the one involved in the notion of standardised implicature or the li(e# %t is 4uite possible that cases "hich are intermediate bet"een purel! conventional and purel! inferential meaning can be usefull! categorized together for theoretical or descriptive purposes# M! argument e&clusivel! concerns the level of the cognitive mechanisms underl!ing pragmatic phenomena# 0pecificall! "hat % do argue is that if one concedes that the recover! of coded meaning is sensitive to the degree of associative strength then conceiving of implicatures in terms of inferential processes ma! cause inconsistenc! < especiall! in le&ical pragmatics < since it is far from clear that the alleged inferential processes can complement the associative recover! as the! should# There is ho"ever a natural "a! to account for implicature'li(e effects "ithout impl!ing an! strong discontinuit! bet"een le&ical retrieval and alleged implicatures# The position adopted b! Recanati in le&ical pragmatics is the closest to adopting this vie" insofar as the ver! same associative mechanism is held to be responsible for both le&ical retrieval and -primar!. pragmatic inferencing# More precisel! in Recanati's vie" primar! pragmatic processes are associative instead of genuinel! inferential but the! have nonetheless inferential'li(e effects# % propose to e&pand this approach be!ond its original boundaries: implicatures in general are reinterpreted as implicature'li(e effects delivered b! associative processes and this ma! account for their gradual decline as the use of "ords becomes more and more > conventionalised# %n practice % "ill first introduce MOR -section 6.# Then % "ill e&amine the arguments that call MOR into 4uestion and the "a! these arguments affect theorizing on G=%s and le&ical pragmatics -section 9.# %n section > % "ill propose an alternative account based on Recanati's non'inferential vie" of le&ical pragmatics# Binall! given relevance theorists' particular vie" of pragmatic inferences % "ill separatel! consider their approach to le&ical pragmatics in the light of the continuum problem -section 8.# 2. The Role of MOR in Grice's Theory Grice's central intuition is that sentences ma! be used to conve! meanings "hich are different from their conventional ones and that this depends on a human inferential abilit! based on rational principles than(s to "hich the addressee ma! recover the further meaning implicated b! the spea(er "hen in a given conte&t she utters a sentence endo"ed "ith a given conventional meaning# :ith a famous e&le -Grice /131: 96. let us consider the follo"ing conversational e&change bet"een t"o interlocutors A and 5 "ith A standing b! an obviousl! immobilized car: -/. A: % am out of petrol# 5: There is a garage round the corner# 5's ans"er implicates that as far as she (no"s A "ill find the garage open and it "ill have petrol# 5ut obviousl! these implicatures are not part of the conventional meaning of 5's sentence: the! are rather suggested b! a consideration of the presumable communicative goal of 5's utterance given both its conversational and ph!sical conte&t# Although protot!pical pragmatic inferences concern < ;ust as in the above e&le < the further conclusions to be dra"n from complete sentential meanings -in the light of the conte&t. Grice has also ta(en into account a number of cases "here pragmatic processes are re4uired in order to determine the meaning that single "ords -or phrases. contribute to complete sentence and utterance meaning# On the one hand these cases include disambiguation of ambiguous "ords and inde&ical reference assignment# $o"ever as =arston -67/7. has correctl! emphasised it seems as if Grice thought of disambiguation and 8 reference assignment in terms of mechanical processes of conte&tual best fit rather than genuine processes of rational inferencing based on a consideration of the spea(er's communicative intention# 5ut there are also cases in "hich according to Grice the contribution of single "ords or phrases to utterance meaning is determined b! means of implicatures that is genuine pragmatic inferences# C&les of this are indefinite descriptions and the counterparts of logical connectives in ordinar! language# Bor instance Grice -/131. invites us to consider an utterance of the follo"ing sentence: -6. A man came to m! office !esterda! afternoon# As a general case "e "ould be surprised to discover that the man "as the spea(er's husband since the indefinite phrase a man suggests that the spea(er is not intimatel! related to that person# Grice's proposal is that this occurs as the result of a specific (ind of pragmatic inference that he calls generalised conversational implicature -G=%s.# %n contrast "ith particularised conversational implicatures "here the implicated meaning is made to depend on the specific conte&t of utterance G=%s are such that the use of a certain form of "ords DEF "ould normall! -in the absence of special circumstances. carr! such'and' such an implicature or t!pe of implicature -Grice /131: 92.# 5! describing the meaning contributed b! the phrase a man as the result of an implicature Grice intends to e&clude that it is part of the fi&ed le&ical meaning of the e&pression* in fact the implicated meaning ma! be cancelled in case the implication is in contrast "ith co'te&tual or conte&tual information# As to logical connectives and their counterparts in ordinar! language let us consider the follo"ing e&le: -9. % too( off m! shoes and got into bed# %n this sentential conte&t the use of and appears to impl! a temporal priorit! "hich is not part of its truth'functional meaning as is sho"n b! the fact that -9. and ->. have the same truth'functional meaning: ->. % got into bed and too( off m! shoes# Again Grice's proposal is that temporal priorit! is pragmaticall! implicated so that "e do not need to ? assume either that and in the ordinar! language has a different meaning than its logical counterpart or that it is ambiguous bet"een t"o -or more. different meanings# Another "ell studied case in point is that of scalar implicatures that is pragmatic effects due to implicational scales# These are sets of le&ical items that ma! be ordered in accordance "ith their informative value so that the use of a less informative item -a "ea(er value. on the scale implies that stronger values on the scale do not appl! -$orn /126 /13>* Gazdar /121.# Bor instance in the sentence: -8. Gohn ate some of the ca(e the presence of some suggests that Gohn did not eat all the ca(e since other"ise the spea(er should have chosen the more informative all# %n sum in all the cases e&emplified in -6.'-8. the claim is that the linguistic forms at issue conve! meanings "hich need not be and in fact are not le&icalised# The! "ould rather be obtained as the result of pragmatic inferences though of a special (ind: implicatures "hich do not depend on specific conte&ts of utterance but are normall! conve!ed b! certain forms of "ords# %t is mostl! "ith regard to this sort of cases that MOR displa!s its function# As far as implications of the standard sort < e&emplified in -/. < are concerned there is a clear intuition that the implicated meaning is different from the conventional one and in fact produced b! means of a pragmatic inference# 9 On the contrar! in the e&les -6.'-8. there is no such clear intuition# On the one hand the allegedl! implicated meanings seem to share "ith le&ical meanings important features < independence from conte&ts of utterance* standard association "ith specific forms of "ords# On the other hand the! are nonetheless claimed to be implicated meanings in that the! share important features "ith protot!pical -i#e# particularised conversational. implicatures: in particular the! appear to be both calculable and cancellable -% "ill turn to calculabilit! belo".# %t is precisel! because the cases in point cannot confidentl! and undoubtedl! be classified either as le&ical meanings or as implicatures since the! e&hibit features of both that Grice feels the need to propose a methodological principle such as MOR# %n the absence of strong evidence supporting one h!pothesis over the other Grice suggests it could be advisable to rel! on a simple principle of parsimon!: "e should refrain from e&plaining some use of an e&pression b! appealing to specific conventions insofar as a more general e&planation < based on general principles of rational 9 Recanati -e#g# 677>: >>. "ith his availabilit! principle and availabilit! condition has importantl! contributed to emphasize the role of this intuition# 2 communication < is available# %n the light of this principle then the cases at issue "ould be better thought of as a special (ind of implicature# Ho" that "e have presented the theoretical bac(ground "hich motivates MOR let us consider some recent ob;ections that have been made to this principle# 3. Arguments Against MOR 5ontl! -6778. has summarised three main traditional arguments against MOR and he has himself pointed to a further difficult! "hich has received little attention to date# Although his real aim is to sho" that those ob;ections can be overcome % intend to suggest that his attempt to defend MOR is fla"ed# The first ob;ection considered b! 5ontl! is that there is at least a class of linguistic phenomena in the light of "hich the commitment of MOR to parsimon! appears misleading since it "ould lead to "rong predictions# The linguistic phenomena at issue are dead metaphors# > %t is generall! held that one' time metaphorical uses can be conventionalised as a conse4uence of repetition so that a "ord form ma! come to have a secondar! meaning that is pragmaticall! predictable and !et full! conventional -5ontl! 6778: 61?.# ,ragmatic predictabilit! is held to be under the name of calculabilit! one of the characterising features of implicatures# :hat MOR prescribes is that "henever a "ord use is calculable it is more parsimonious to conclude that it is in fact calculated rather than le&icall! recovered# $o"ever dead metaphors are clear countere&les to this conclusion since ever!one agrees that the! have conventional meanings "hich can be le&icall! recovered in spite of the fact that those meanings are in principle pragmaticall! calculable# Therefore parsimonious as it can be the inference from being calculable to being calculated does not hold in general: "ord uses can be calculable and nonetheless conventionalised# :hile this first argument concedes that it is parsimonious to infer that calculable meanings are also calculated though raising the ob;ection that this inference does not hold in general the second argument calls into 4uestion the ver! notion of parsimon! at issue: implicature accounts loo( at least superficiall! to multipl! inferential labor DEF# Thus there are tradeoffs involved and the account "hich > 5ontl! observes that versions of the argument from dead metaphors have been proposed b! :al(er -/128. Morgan -/123. 0adoc( -/123. Reimer -/113. and @evitt -677>.# 0ee also @evitt -6772.# 3 is semanticall! more parsimonious ma! be less parsimonious all things considered -5ontl! 6778: 612.# :hat is at issue here is cognitive not epistemological parsimon!# 0pecificall! the point is that ;ust as cognitive linguists have repeatedl! pointed out cognitive parsimon! on the storage side is inversel! correlated "ith that on the processing side -5arsalou /116: /37'/* =roft and =ruse 677>: 623* Goldberg /118: 2>.# This consideration has been ta(en as supportive of the vie" that information can be redundantl! stored at different levels of linguistic ta&onomies so that it can be directl! recovered from linguistic forms at an! level rather than inferentiall! derived from superordinate levels# Although at first sight redundant storage might not appear a rational option for cognitive s!stems the possibilit! to recover information non'inferentiall! might be instead a valuable gain on the side of processing effort# %n the same line b! coding multiple meanings rather than emplo!ing pragmatic implicatures one could get parsimon! on the processing side# To be more precise 5ontl! -6778: 612. observes that postulating an ambiguit! might not dispense "ith inferences entirel!# 8 %nsofar as there are multiple meanings the addressee needs to ta(e into consideration conte&tual clues in order to determine "hich of those meanings is actuall! intended in a circumstance# Ho" if selection of conte&tuall! relevant meanings re4uires inferential processing then conventionalising a specific use of a "ord ma! lo"er the comple&it! of inferential processing but not remove it entirel!# 5e that as it ma! even in this case parsimon! in storage and in processing are inversel! correlated and then the argument from parsimon! of storage is far from being conclusive# The third argument bears upon the epistemological status of MOR proper that is its status as a methodological device to select bet"een competing h!potheses# The point is that methodological virtues in themselves have little impact on genuine ontological issues# On an ontological ground the burden of proof is probabl! on the enemies of ambiguit!: Given that ambiguit! is hardl! rare then one "onders "hether a semantic theor! ought reall! to minimize it DEF: perhaps "e ought to assume -ceteris paribus. that every regular use of an e&pression represents a special sense -5ontl! 6778: 611'977 his emphasis.# %n his attempt to sho" that MOR can be defended from the received criticisms 5ontl! starts from this last epistemological point# $e ma(es an appeal to 0ober's -/133. claim that arguments from parsimon! ma! be cogent after all provided that the! are emplo!ed not as domain'general principles of 8 $ere and else"here 5ontl!'s use of the term ambiguit! is potentiall! misleading# A "ord is said to be ambiguous "hen it has a pluralit! of possibl! unrelated meanings "hile the debate on MOR concerns cases in "hich meanings can be rationall! inferred from one another and are therefore related to each other# These are better referred to as cases of pol!sem!# 1 scientific reasoning but instead "ith reference to domain'specific assumptions about the processes that generate the phenomena under stud!# %n practice 5ontl!'s argument is that there is a tendenc! to parsimon! that is specific to semantic ac4uisition# %n fact it is a "ell established ps!chological observation that children have difficulties learning homon!ms ? and this suggests that language ac4uisition is semanticall! conservative: children will posit new meanings for familiar words only when necessary -5ontl! 6778: 976 emphasis mine.# This conclusion in 5ontl!'s opinion is precisel! the sort of domain'specific assumption that ma(es MOR a reasonable principle for theor!'choice in semantics: since sub;ects resist positing ne" meanings -e&cept "hen necessar!. then there is a specific cognitive ;ustification for semantic parsimon!# Infortunatel! the argument cruciall! depends on ho" one interprets the "ords "hen necessar!# 2 5ontl! apparentl! assumes that the right interpretation is: "hen there is no pragmatic inference b! "hich an addressee could calculate the intended meaning or more s!ntheticall! "hen meanings are not calculable# Onl! b! this interpretation "ould the argument from ac4uisition spea( in favour of MOR: if sub;ects resist positing ne" meanings unless these cannot be calculated then "e should not attribute stored meanings to sub;ects "hen those meanings are calculable# 5ut there are other possible interpretations of "hen necessar! for instance "hen forced b! repetition# %n this case observations from ac4uisition "ould spea( in favour of a "holl! different conclusion that is the conclusion that "e should better not attribute stored meanings unless "hen those meanings are repeatedl! used -in association "ith a phonological form.# 0ince 5ontl! simpl! assumes the former interpretation "ithout providing an! reason for it his argument is a petitio principii: that ne" meanings are resisted -and eventuall! not stored. in case they are calculable is "hat should be sho"n "hile it is in fact tacitl! ? The same terminological "arning of note 8 applies here to the term homon!m: this specificall! refers to multiple meanings "hich are unrelated to each other# 5ontl! apparentl! thin(s that the cited evidence licenses conclusions concerning pol!sem! as "ell# As noted b! an anon!mous referee this inference is far from "arranted: children could resist homon!m! and !et be perfectl! happ! "ith "ords being pol!semous# Although this is in itself a serious problem for 5ontl!'s argument % propose a more general criticism "hich applies also in case it "ould turn out that children resist pol!sem! as "ell# 2 The evidence cited b! 5ontl! for instance Mazzocco -/112. onl! sho"s that !oung children -9' to 8'!ear'olds. tend not to use alread! (no"n "ords for ne" referents "hile the! can easil! learn ne" "ords for those referents# %n the light of those studies no conclusion can be dra"n as to the reasons "h! children are so parsimonious in the former case or the conditions at "hich such parsimonious tendenc! "ould be overridden# %n particular there is nothing suggesting that children are al"a!s and onl! parsimonious "hen meaning can be calculated# /7 assumed b! the argument# Hot onl! "ould 5ontl! need an argument in order to adopt his interpretation of "hen necessar! "ithout begging the 4uestion but he also needs a "a! to manage the case of dead metaphors# As "e sa" dead metaphors seem to sho" that meanings can be stored even "hen the! are calculable and this seems to be a strong case against the thesis that children in particular and people in general do not store ne" meanings e&cept "hen these are not calculable# %n fact 5ontl! attempts to provide an argument for e&plaining a"a! the case of dead metaphors# On the basis of his developmental account of parsimon! 5ontl! proposes the follo"ing anal!sis of "hat he assumes to be an instance of dead metaphor that is the metaphorical use of incense: %t ma! be therefore that children at the relevant developmental stage DEF lac( the understanding necessar! to predict that incense could be used to mean to make or become angry# The! might not realize for instance that incense means an aromatic substance that burns with a pleasant odor and even those "ho do probabl! lac( the general bac(ground (no"ledge necessar! to appreciate the metaphorical connections bet"een burning and emotion# Cither "a! the metaphor "ould be dead to children forcing them to learn that use the same "a! the! learn an! arbitrar! convention# -5ontl! 6778: 971. This seems to be a reasonable anal!sis of ho" the "ord incense is ac4uired b! children# 5ontl! apparentl! thin(s that a general lesson can be dra"n from this e&le and it is that children are forced to store metaphorical meanings onl! insofar as the! cannot calculate metaphors for lac( of the (no"ledge re4uired# $o"ever one "onders "hether his anal!sis generalises to an! case of presumed dead metaphors: does it appl! for instance to legs of tables or to hiding behind someone else -in its figurative meaning. and so on and so forthA 0ince there are a number of cases in "hich children appear to have the bac(ground (no"ledge necessar! to appreciate the relevant metaphorical connections then at least in those cases there is still an argument against MOR and for storage of calculable meanings unless one gives independent reasons for believing that metaphorical meanings are stored in incense cases but not in leg of the table ones# 0ince on the contrar! 5ontl! gives no such reason he is again begging the 4uestion# To sum up there is circularit! both in 5ontl!'s developmental argument for MOR and in his specific appeal to the same argument in order to manage the case of dead metaphors# 5ut 5ontl! has also made an interesting observation that helps to shed light on "here the real problem "ith MOR lies# As "e sa" there is a third traditional argument against MOR -besides the argument from dead metaphors and the argument from the distinction bet"een epistemolog! and ontolog!.: the argument of // tradeoffs bet"een storage and processing parsimon!# %nferential processes are thought to be cognitivel! more comple& and in this sense less parsimonious than simple disambiguation# Ho" 5ontl! recalls that there is a standard response to this ob;ection "ithin the Gricean literature: specificall! "ith regard to G=%s it is claimed that implicature calculation is largel! unconscious and implicit: bac(ground assumptions can be ta(en for granted steps can be s(ipped and onl! rarel! need the entire process breach the surface of consciousness -5ontl! 6778: 972.# As a conse4uence it is held that implicatures of this sort are no more difficult than disambiguation# $o"ever 5ontl! sees in this argument an interesting difficult! that -to m! (no"ledge. Griceans have never ade4uatel! addressed for it is no" 4uite unclear "h! this default interpretation should be considered an implicature rather than an additional sense of the e&pression# To sa! that it is a default interpretation is after all to sa! that spea(ers learn to associate that interpretation "ith the t!pe of e&pression in 4uestion DEF# 0hort'circuited' implicature'calculation is thus hard to differentiate from disambiguation ma(ing the Gricean h!pothesis loo( more li(e a notional variant than a real competitor to the ambiguit! h!pothesis# -5ontl! 6778: 973. 3 This said 5ontl! attempts to get rid of the difficult! b! observing that conventional meanings are inherentl! arbitrar! "hile implicated meanings are not: Bamiliarit! "ith Da givenF use DEF can remove the need to go through the canonical inference but it doesn't change the fact that the use has a 'natural' -i#e# nonconventional. e&planation# %t doesn't change the fact that it is calculable -5ontl! 6778: 973 his emphasis.# %n other "ords the argument goes although familiarit! can cause implicatures to be short'circuited to the point that the! become hard to differentiate from disambiguation nonetheless short' circuited inferences are different from conventional meanings in that the former but not the latter are calculable# Therefore 5ontl! suggests short'circuited implicatures can ensure processing parsimon! "ithout becoming indistinguishable from conventional meanings# 5ut again this move ;ust amounts to ma(ing an appeal to "hat 5ontl! should prove: that if a meaning is calculable then one is ;ustified to assume that it is not conventional# To put it differentl! 5ontl! ma(es a good point "hen he observes that from a cognitive point of vie" short'circuiting of implicatures loo(s prett! much li(e the storage of associative relationships# %f one nonetheless resists considering short'circuited implicatures as a sort of conventionalised meanings for the reason that the former -but not the latter. are calculable then in practice she is ma(ing an appeal to MOR < to the principle that "e should not posit ne" conventional meanings "henever the! are calculable# The 3 5ut see note > above: "hat 5ontl! here calls ambiguit! h!pothesis should be rather called pol!sem! h!pothesis# /6 circularit! pops up again# %t could be argued that there is no circularit! if instead of ma(ing an appeal to MOR as an epistemological principle 5ontl! "ere simpl! pointing at an ob;ective though dispositional difference bet"een mere conventional meanings and short'circuited implicatures# 5ach -/113 "hich revisited 5ach /118* see also Morgan /123. has seemingl! made a similar point# As far as standardised implicatures are concerned he sa!s DtFhe inference is compressed b! precedent# 5ut "ere there no such precedent in "hich case a more elaborate inference "ould be re4uired there "ould still be enough conte&tual information available to the hearer for figuring out "hat is being conve!ed# That is "h! special conventions are not needed for these special cases# -5ach /113: 2/9. 5ach seems to be suggesting here that short'circuited implicatures are distinguishable from genuine conventions not in terms of the actual processes involved but instead in terms of a dispositional propert!: the! could be substituted for b! more elaborate inferences in case it "ere re4uired# $o"ever as a matter of fact more elaborate inferences are normall! not re4uired to the e&tent that the! are compressed b! precedents# Therefore from a cognitive perspective one ma! doubt that such a dispositional propert! has an! import -in the normal case.# Jet me insist on this point# 5ontl! has identified a genuine difficult! in dra"ing a cognitive distinction bet"een short'circuited implicatures and conventionalised interpretations# To the e&tent that the! are compressed b! precedents inferences seem to be replaced b! more straightfor"ard associative processes and intended meanings "ind up being conventional# The fact that intended meanings could still be obtained b! means of more elaborate inferences "ere the! not associativel! recovered does not change associative recover! into something different# As Mazzone -67/9 a. puts it the mere fact that "hen needed a genuine inferential process ma! occasionall! occur as a substitute for associative access does not change the latter into a half'inferential process# %n sum there seems to be a real difficult! for the notion of standardised implicatures: once it is recognised that implicatures can be standardised through repetition then it is far from clear that the! remain implicatures nonetheless# A "holl! standardised implicature is presumabl! changed into a conventional meaning proper# A corollar! of this observation is that since conventionalisation is not an all'or'nothing affair then "e ma! reasonabl! find cases "here meanings are ;ust partiall! conventionalised# ,hillips -67//. has provided an epistemological argument against MOR "hich is based precisel! on this premise: /9 "e can imagine a sorites series going from uses of e&pressions that are determinatel! not conventional -e#g# obvious implicatures. through those regularities in the borderline region that are neither determinatel! conventional not determinatel! not conventional because the e&tra ingredients that are re4uired for a convention to be in place are neither determinatel! present not determinatel! not present to those regularities that are clearl! conventional# -,hillips 67//: ?'2. According to ,hillips the e&istence of a borderline region of indeterminac! casts some doubt on the overall epistemological strateg! pursued b! MOR# This strateg! is in fact based on t"o assumptions: -/. that the linguistic data are indecisive bet"een the implicature account and the pol!sem! account and -6. that these t"o accounts are the best available e&planation of the data# Onl! once these assumptions are made can one appeal to some epistemological principle in order to favour one account over the other# $o"ever as ,hillips observes "hile -/. is clearl! true -6. is probabl! false# There is in fact an alternative e&planation of the data "hich is better than both the implicature and the pol!sem! account: if the data are indecisive bet"een the t"o given that conventionalit! comes in degrees then the most plausible e&planation is that the cases at issue are some"hat intermediate bet"een being "holl! conventional and being pragmaticall! calculated# Hot onl! does this line of thought pose a problem for MOR as an argument for the implicature account it also calls our attention to a fact that cannot easil! be accommodated b! that account# As far as "ord uses are concerned Griceans aim to give a pragmatic e&planation < based on the notion of implicature < in an! case e&cept those for "hich ever!one agrees that a conventionalisation has occurred# Moreover in intermediate cases man! Griceans are read! to appl! the h!brid notion of standardised implicatures that is implicatures that are compressed b! precedents# $o"ever such a notion ma(es little sense from a cognitive point of vie": as "e sa" above it is disputable that a "holl! short'circuited implicature is still an implicature rather than an additional sense of the e&pression -5ontl! 6778: 973.# Therefore the Gricean e&planation for intermediate cases collapses into the pol!sem! account# 5ut this difficult! in establishing a notion that is intermediate bet"een full'fledged implicatures and full! conventionalised meanings also has conse4uences for the notion of implicature as such# 0ince conventionalisation comes in degrees -as emphasised b! ,hillips 67//. one ma! "onder "hat happens to implicatures "hen implicated uses become more and more conventionalised# @o implicatures decline complementaril! in a graded fashionA % (no" of no attempt to provide an ans"er to this issue "hich is "hat % dubbed above the continuum problem for implicatures# This is a problem % maintain for an! theor! of le&ical pragmatics that is based on the notion of /> implicature# 1 A sensible account of ho" "ord meanings are stored and recovered must plausibl! assign some role to associative strength: "ord meanings can be more or less conventionalised as a function of regularit! of use# As a conse4uence if implicatures have to complement conventionalisation the! should decline graduall! as "ell# 5ut then "e apparentl! need an account of implicature'li(e effects "hich does not impl! an! strong discontinuit! bet"een le&ical retrieval and alleged implicatures# %n the follo"ing section % "ill s(etch the general lines of such an account# 4. e!ical "ragmatics #ithout Inferences As % said above it is probabl! a main tenet of most current pragmatic theories inspired b! Grice's "or( that implicatures have a strictl! inferential nature in the sense that the! involve some sort of -possibl! implicit. reasoning instead of mere associative recover! of coded information# Therefore pragmatic processing is generall! conceived of as an inferential affair "ith the important e&ception of Recanati's -677>. account of le&ical pragmatics# $e proposes that the determination of e&plicit meaning although being a pragmatic affair is nonetheless ensured b! "holl! associative processes to "hich he refers as primar! pragmatic processes# 5! contrast Recanati maintains that the transition from the e&plicit meaning of the utterance < fed b! primar! pragmatic processes < to its implicit meaning is ensured b! secondar! pragmatic processes conceived in terms of genuinel! inferential personal'level processes# $o"ever in Recanati's opinion primar! pragmatic processes are apt to mimic genuine pragmatic inferences -Recanati 6772.# Therefore in his vie" of le&ical pragmatics the same associative mechanism is responsible for both le&ical retrieval and implicature'li(e effects# %n "hat follo"s % am going to s(etch a more e&plicit cognitive interpretation of this vie" -along the lines of Mazzone 67//. than the one provided b! Recanati himself# Bor the sa(e of clarit! t"o main components must be distinguished in the account % propose# The first is the spreading of activation from linguistic e&pressions to associativel! related representations "ithin the conceptual net"or(* the second component is "hat Recanati calls the accessibilit! shift 1 More generall! it is a problem for an! theor! of le&ical pragmatics that is based on inferential mechanisms possibl! including e&plicatures in Relevance Theor!# $o"ever since Relevance Theor! sensibl! differs from the approaches considered so far < it conceives of pragmatic inferences in terms of specialized automatic processes and it does not ma(e an! use of the notion of standardised implicature < % "ill address it separatel! in section 8# /8 occurring in the course of processing as soon as ne" information is made available to the cognitive s!stem# The first mechanism is essentiall! dependent on -provisionall!. fi&ed relationships of accessibilit! based on previousl! stored regularities# %n practice a linguistic e&pression can be associated "ith different strengths to a variet! of concepts# More generall! an! informational input ma! have a set of representations associated to it "hich are therefore accessible through spreading activation* the differential strength of those associations determines the order of accessibilit! for the representations in that set < that is the order in "hich the activation spreads from a given informational item to the related representations# This temporal order of accessibilit! is correlated to ho" strongl! each representation in the set is activated as a conse4uence of the activation of the informational item: representations that are accessed first are also -ceteris paribus. activated at the highest degree since activation declines as it spreads# The order of accessibilit! rooted in stored associations ho"ever cannot be the last "ord# %ncoming information is usuall! processed "ithin the conte&t of both previous patterns of activation in the brain and immediatel! subse4uent trains of activation due to further information# This ma! e&plain "hat Recanati calls accessibilit! shift in the course of processing# Representations that "ould have received the highest activation in the absence of an! conte&t can be out"eighed b! others in case supplementar! information is provided# More generall! conte&tual information ma! enhance or reduce the activations induced b! an isolated item thus subverting the standard pattern of accessibilit!# A crucial component in Recanati's account is the notion of schemata bridging informational items# A schema can be described as a representation t!ing t"o or more concepts together into a structured "hole# %n practice an! informational item ma! activate schemata "hich in turn spread the activation to their other components# 0chemata are appealed to b! Recanati in order to e&plain ho" accessibilit! shifts ma! promote the search for coherence in interpretation# %n his account of primar! pragmatic processes the initial advantage of a given interpretation ma! be overridden due to the fact that a less accessible meaning has a better fit "ith < and therefore receives further activation b! < some informational item via schemata of "hich the! are both components# %n this "a! meanings that are more coherent "ith other informational items in the conte&t come to be preferred over less coherent ones# There are t"o plausible assumptions underl!ing this account: first that conceptual information is constituted of concepts multipl! related to each other through schemata "hich specif! the nature of their relationships* /? and second that schemata are actual components of the conceptual net"or( and therefore activation spread through them as "ell as through the concepts the! tie together# :h! does Recanati hold that the processes described above are apt to mimic genuine inferential processesA The reason is that once the role of accessibilit! shift and schemata is put into the picture associative spreading activation is far from appearing as a random process# Kuite on the contrar! it performs so as to obtain the same motivated results as if conclusions "ere dra"n b! "a! of rational inferential processes# Jet us consider the follo"ing e&le from =arston -6772.: -?. Ann: % e&pected Gane to be here b! no"# 5ob: 0he missed her coach# The "ord form coach is potentiall! ambiguous bet"een t"o possible meanings: =OA=$ / L %H0TRI=TOR and =OA=$ 6 L 5I0# =arston's -6772. account of ho" the addressee ma! come to conclude that coach refers to a bus is along the usual lines of RT# 0pecificall! she appeals to an inferential process "here certain conclusions are dra"n from certain assumptions and in particular the crucial assumption is the follo"ing generalisation from e&perience: -2. M%00%HG A @C0%GHATC@ =OA=$ 6 %0 A RCA0OH BOR A ,CR0OH HOT ARR%M%HG :$CH CN,C=TC@# $o"ever this assumption is nothing but a schema in Recanati's sense that is a pattern connecting concepts through some specific (ind of relation# /7 $ere the relation at issue is one of causal e&planation+rational ;ustification: a person might not arrive "hen e&pected for the reason that s+he has missed the bus# The schema licenses the conclusion that presumabl! 5ob means that Gane has missed a bus since this interpretation ma(es his utterance conte&tuall! relevant b! giving a reason for Gane's dela!# 5ut that conclusion "ould be e4uall! prompted b! an associative search for coherence driven b! the schema at issue# Marious "ord forms in Ann's and 5ob's utterances -e&pected missed coach. ma! /7 Recanati -677>: 92. emplo!s the e&le of the causal schema STEAL (X ! "S A##ESTE$ (X in order to e&plain the interpretation of the pronoun in the sentence Gohn "as arrested b! a policeman !esterda!* he had ;ust stolen a "allet# As to -2. it is conceptuall! e4uivalent to the causal schema %"SS T&E '(A'& (X ! $( )(T A##"*E +&E) EX,E'TE$ (X# /2 be e&pected to spread activation to the schema insofar as the associated concepts -TO CN,C=T TO M%00 =OA=$ 6 . are component of the schema and this in turn "ould enhance the activation of the conte&tuall! relevant meaning of coach# %n sum "hat guarantees the conclusion is not the fact that the schema is emplo!ed in a genuinel! inferential process but rather the schematic information in itself due to its po"er to constrain and drive the interpretation# %t is such a schematic organisation of conceptual net"or(s that ma(es it possible for associative processes to mimic genuine inferential ones# % have argued for this vie" at more length in Mazzone -67//.# :hat % "ant to emphasise here is that in this account there is no discontinuit! bet"een le&ical retrieval and inferential effects# 0chemata can constrain utterance interpretation b! "a! of the ver! same mechanism involved in le&ical retrieval that is the d!namics of activation the! ta(e part in# // To be sure in the above e&le "e assumed that the intended meaning is alread! conventionalised and therefore encoded b! the spea(ers involved: it is one of the le&ical meanings of the ambiguous "ord form coach# $o"ever as "e observed in the third section ambiguit! -and pol!sem! // An anon!mous referee has urged that % should ma(e clear "hat -ho" much. is encompassed b! the notion of schema insofar as m! account places huge "eight on it# The concern is that the notion cannot be broad enough to cover the entire contribution of enc!clopedic+general (no"ledge to utterance understanding# Burthermore the same referee "onders "hether m! claim of a continuit! bet"een le&ical retrieval and inferential effects is compatible "ith e&perimental "or( such as that of Rubio Bernandez -6772. "hich seems to sho" that meaning selection and meaning construction are different processes# % cannot full! ans"er these 4uestions here# Jet me ;ust sa! this much# As to the first issue % adopt a ver! general notion of schema according to "hich an! pattern of regular covariation bet"een t"o or more concepts is a schema# % believe that this notion is ;ustified b! its role in e&plaining ho" concepts are both formed -Mazzone in preparation. and used in cognitive processing -Mazzone in press.# As to the other issue % do not claim that there is no discontinuit! at all bet"een meaning selection and meaning construction# 0pecificall! % do not den! that meaning construction might involve "or(ing memor! and attentional processes# On the contrar! in Mazzone -67/9 b c. % e&plicitl! maintain that for a full e&planation of utterance understanding associative processes need to be complemented b! attentional resources# $o"ever this must be compatible "ith the fact underl!ing the continuum problem that is the fact that pragmatic inferences must decline progressivel! to the e&tent that uses become more and more conventionalised# This is easil! e&plained if "e assume that the same associative d!namic is involved in the t"o cases: the indirect associative relation bet"een t"o -or more. concepts ma! be graduall! short'circuited b! repetition and this in turn affects ho" those concepts ma! activate each other < -more or less. indirectl! via schemata and -more or less. directl!# %ncidentall! the opposition bet"een continuit! and discontinuit! in cognitive processing might not be as neat as it is often assumed -see also the end of section 8* more on this in Mazzone 67/9 c.# /3 as "ell. does not dispense "ith inferences or inference'li(e effects since in order for the appropriate meaning to be selected the conte&t needs to be ta(en into account an!"a! and schematic assumptions are needed to bridge intended meaning and conte&tual items# A different case is "hen the intended meaning is not encoded b! the hearer# %n this case the recover! of the meaning is more indirect# Jet us suppose for instance that the listener has not -!et. encoded the concept 5I0 as a conventional meaning of the "ord form coach# :e might nonetheless assume that her conceptual s!stem contains the information e&pressed b! the schema -2. although not in that precise form since b! h!pothesis that s!stem does not include a concept =OA=$ 6 -L 5I0. associated to the "ord form coach# Then the schema "ould rather be as follo"s: -3. M%00%HG A @C0%GHATC@ 5I0 %0 A RCA0OH BOR A ,CR0OH HOT ARR%M%HG :$CH CN,C=TC@ Cven in this case the schema could help the listener to identif! 5I0 as the intended meaning of coach though in a more indirect "a!# The schema could be "ea(l! activated < together "ith other schemata providing alternative possible reasons "h! a person could not arrive "hen e&pected < b! Ann's utterance -% e&pected Gane to be here b! no".# Burther the "ord form missed in 5ob's ans"er "ould probabl! add activation to a subset of those schemata specificall! the ones involving missing buses trains etc# Then the conte&t -or stored (no"ledge. might provide further clues leading the listener to select M%00%HG T$C 5I0 amongst the possible reasons# %n sum it is not inconsistent to thin( that insofar as the intended meaning is recovered at all /6 this could be ensured again b! the activation of schemata bridging conte&tual clues -or stored (no"ledge. and the intended concept# 5esides in a number of cases it happens that the le&icalised meaning of a "ord form and the meaning attributed to it in a circumstance are tied b! some sort of relation -meton!m! similarit! etc#.# %n those cases although the concept "hich is provided b! the schema is not itself a le&icalised meaning of /6 An anon!mous referee has suggested that in this case confusion or misinterpretation and not the recover! of the "ord meaning is the most li(el! outcome: the hearer "ould tr! to find some "a! of ma(ing sense of the proposition that Gane missed her instructor and that this has someho" dela!ed her arrival# % do not argue "ith this suggestion# M! aim is ;ust to s(etch the general lines of ho" even the most difficult case < that is the case in "hich the le&ical meaning is un(no"n to the hearer < can be accounted for in terms of activation of schemata# 5ut this does not commit me to the claim that the hearer necessaril! succeeds in recovering the meaning# /1 the "ord form there is presumabl! some further schema -be it meton!mic metaphoric etc#. connecting the le&ical meaning and the intended one# These are clear cases in "hich the intended meaning is calculable that is it can be derived from the le&icalised one in a motivated manner and in "hich pol!sem! -versus homon!m!. ma! result "hen the intended meaning becomes le&icalised# The crucial point is that the proposed account allo"s for a gradual transition bet"een the cases considered above due to the fact that le&ical retrieval and the contribution of schemata are ;ust different facets of the same activation d!namics# Thus "hen the intended meaning is not le&icalised so that some "rong concept -or no concept at all. tends to be activated b! simple le&ical retrieval associative activation of schemata ma! be sufficient to !ield the re4uired accessibilit! shift and drive the search for the right concept# %n the end the intended meaning ma! become progressivel! conventionalised: this e4uates to the formation and strengthening of a direct connection bet"een the "ord form and the meaning so that the more indirect route mediated b! other schemata is short'circuited and eventuall! b!passed# 0ince in this vie" the indirect route does not involve genuine inferences but rather associative mechanisms mimic(ing inferences the gradual decline of this route as a conse4uence of conventionalisation does not raise an! difficult!# %t should be noticed that this approach is not incompatible "ith the intuition according to "hich at least in some cases "ord uses that are calculable might not be -much. conventionalised in spite of their fre4uenc!# The point is that the fre4uenc! "ith "hich a "ord form conve!s a meaning is ;ust one factor of its conventionalisation# Other plausible factors are the fre4uenc! of previousl! le&icalised meanings the absence of s!non!ms the independence of ne" uses from old meanings the determinac! of "ord uses the grammatical categor! < and possibl! others# Jet us consider for instance the metaphorical sentence -1. Guliet is the sun# =ould the phrase the sun come to be le&icall! associated to < let us sa! < the concept 5CAIT%BIJ even in case it "ere fre4uentl! used in that senseA The ans"er must be mostl! negative for in this case almost all the factors cited above disfavour le&icalisation# The "ord form sun is basicall! used to e&press the concept 0IH -as to the factor fre-uency of previously le.icalised meanings.* the concept 5CAIT%BIJ is alread! le&icalised in Cnglish -presence of synonyms.* the phrase the sun in conte&ts such as -1. has a much more indeterminate meaning than 5CAIT%BIJ -determinacy of word uses.* the 67 phrase the sun belongs to a different grammatical categor! than the meaning 5CAIT%BIJ "ould re4uire# Bor another e&le let us consider the meton!mic sentence -discussed in Hunberg /122.: -/7. The ham'sand"ich left "ithout pa!ing# The phrase the ham'sand"ich is intended as referring to the person "ho ordered the sand"ich than(s to a transfer from one component to another of a cafO+restaurant schema# Ho" it is not onl! that the literal meaning is no less -and probabl! more. fre4uent than the figural one -fre-uency of previously le.icalised meanings.* moreover this case is so construed that the literal meaning is salient in the conte&t and presumabl! has a role in retrieving the figural meaning -there is little independence of new uses from old meanings.# Bre4uenc! presumabl! favours le&icalisation of secondar! meanings in inverse proportion to ho" much the conte&ts of use evo(e primar! meanings# %n sum in both the e&les above < -1. and -/7. < one should not e&pect a significant amount of conventionalisation# 5ut this is not evidence that "henever uses are calculable then the! are calculated -b! "a! of genuine inferences. instead of conventionalised# As a matter of fact uses that are calculable are nonetheless conventionalised in man! other cases: "e have mentioned dead metaphors but this also seems to appl! to dead meton!mies such as :hite $ouse used to refer to the ,resident of the Inited 0tates and so on and so forth# :hether calculable meanings are conventionalised or not depends on fre4uenc! of "ord use together "ith the interacting factors "e considered above# Given that b! this account various factors of conventionalisation < including the fre4uenc! "ith "hich a "ord form conve!s a given meaning and the fre4uenc! of previousl! le&icalised meanings < have continuous values the overall picture is still compatible "ith the idea that there is a continuum bet"een "holl! conventionalised and non'conventionalised uses and that conse4uentl! a gradualist account is needed at the cognitive level# /9 $. Rele%ance Theory and the &ontinuum "ro'lem /9 Remember that strictl! spea(ing % am not concerned here "ith conventions per se but "ith ho" conventions are implemented in individual cognition# 6/ %n vie" of our above considerations it is interesting to briefl! consider relevance theorists' position in le&ical pragmatics# %nsofar as the! too insist on the genuinel! inferential nature of pragmatic processing their position might seem to be plagued b! the continuum problem ;ust as much as the more traditional Gricean accounts# 0ome aspects of their approach ho"ever suggest the possibilit! of a different interpretation "hich is more in line "ith m! proposal in the previous section# To start "ith :ilson and =arston -677? 6772. have recognised that associative processes ma! have a role to pla! in le&ical pragmatics -more on this point in Mazzone 67//: section 9.# /> Associations "ould determine the initial accessibilit! of concepts and therefore the initial probabilit! that a concept is a component of the e&plicit meaning of utterances although this meaning is eventuall! ad;udicated b! the properl! inferential component of the process# C&plicit meaning and conte&tual assumptions are in fact held to be the premises from "hich conte&tual implications -implicated conclusions. have to be inferred# Relevance theorists ho"ever have a non'linear conception of this inferential process# The! spea( of a mutual ad;ustment mechanism "or(ing in parallel "ithout a rigid se4uential order# This means that the premises of the inference might be recovered or at least ad;usted "ith the help of bac("ard inferences from the conclusion in case the conclusion is made independentl! available than(s to conte&tual clues# 0pecificall! the h!potheses about e&plicit content -and the concepts that compose it. can be revised as a function of both the conte&tual assumptions and the conclusion-s. involved# As an e&le consider the follo"ing e&change -from :ilson and =arston 677?: >66. "here the first spea(er as(s the 4uestion -//a. about the addressee's !ounger sister =aroline "ho is manifestl! not a princess and then the addressee ans"ers as in -//b.: -//. a# :ill =aroline help us clear up the flood damageA b# =aroline is a princess# The meaning of princess in -//b. has to be modulated in accordance "ith the conte&t so as to conve! the concept ,R%H=C00P "hich differs from the le&icalised concept ,R%H=C00# 0pecificall! :ilson and /> Bor instance :ilson and =arston -677?: >61. claim that associative lin(s ma! affect the outcome of the mutual ad;ustment process b! altering the accessibilit! of conte&tual assumptions and implications but the resulting overall interpretation "ill onl! be accepted as the spea(erQs intended meaning if it satisfies the hearerQs e&pectations of relevance and is properl! "arranted b! the inferential comprehension heuristic discussed Dabove in that paperFQQ# 66 =arston propose that the follo"ing inference is dra"n: -/6. a# Explicit content: =AROJ%HC %0 A ,R%H=C00P b# Contextual assumption: A ,R%H=C00P @OC0H'T =JCAR I, BJOO@ @AMAGC c# Contextual implication: =AROJ%HC N :OH'T $CJ, I0 =JCAR T$C BJOO@ @AMAGC :ilson and =arston observe that the conte&tual assumption in -/6b. is unli(el! to be stored in the enc!clopedic entr! for ,R%H=C00 but the! suggest that this feature should be straightfor"ardl! derivable in the course of the mutual ad;ustment process b! a combination of for"ard inference from e&isting enc!clopaedic features -e#g# IHI0C@ TO ,CRBORM%HG MCH%AJ TA0)0 IHA==I0TOMC@ TO MAHIAJ JA5OIR. and bac("ard inference based on the e&pected t!pe of conclusion in D-/6=.F -:ilson and =arston 677?: >66.# =learl! the inferential process is not ta(en here to occur e&clusivel! in a for"ard direction from the premises to the conclusion# On the contrar! there is a specific e&pectation about ho" the utterance in -//b. is li(el! to achieve relevance as an ans"er to -//a. -cfr# :ilson 677>: 98>. and therefore there is some e&pectation about the most li(el! conte&tual conclusion# This conclusion in turn contributes to the recover! of the premises < specificall! the conte&tual assumption -/6b. and the concept ,R%H=C00P in the e&plicit content -/6a.# 0uch a mutual ad;ustment is e&actl! "hat one "ould e&pect in case the process at issue "ere associative# %n associative processes different representations ma! act on each other as mutual constraints independentl! from their order of activation# Then one "ould be tempted to conceive of the described process as inferential in the limited sense that there are schemata connecting the involved representations so as to potentiall! form a ;ustificator! structure < that is a structure that could be appealed to in order to inferentiall! ;ustif! the conclusion < although schemata and representations act on each other as associative constraints not literall! as se4uential steps of an inferential derivation# %n this case "e could claim that associative processes mimic inferential ones ;ust as Recanati -6772. suggested# An interpretation of RT along these lines seems consistent "ith the follo"ing claim made b! :ilson and =arston -6772: 6>>.: Brom a cognitive point of vie" all inferential relationships are also associations: an inferential mechanism establishes s!stematic correspondences bet"een -constituents of. premises and -constituents of. conclusions -see also 0perber and :ilson 6773.# 69 %n sum one could "ea(en the thesis of genuinel! inferential processes to the point of sa!ing that there are inferences provided that ;ustificator! structures are involved "hatever the cognitive mechanisms b! "hich this occurs# Of course % "ould agree that pragmatic processes are inferential in this specific sense# $o"ever relevance theorists "ould hardl! be content "ith the idea that inferential pragmatic processes can be accounted for in "holl! associative terms# :hile the! clearl! concede that the associative d!namic of activations is responsible for feeding concepts into the e&plicature -and also for feeding conte&tual assumptions into the derivation. the! seem to assume that the inferential stage of processing re4uires a cognitive process of a 4uite different sort than the associative one# %n m! opinion though this assumption of a cognitive discontinuit! might not be eas! to reconcile "ith the RT's frame"or( for le&ical pragmatics described above for three reasons# Birst "ithin that frame"or( all the inferential steps need to be open to the operations of associative processes# %f inferences "ere thought of as se4uential derivations based on a set of rules then the conclusions could be thought to simpl! follo" once the premises are set# 5ut given the h!pothesis of a mutual ad;ustment process even conclusions must be associativel! sensitive to conte&tual inputs in much the same "a! as e&plicit content and conte&tual assumptions are# Onl! if conclusions are independentl! activated b! the conte&t can the! license bac("ard inferences driven b! specific e&pectations of relevance# Moreover not onl! can an! of the inferential steps be -provisionall!. determined b! conte&tual items b! means of associative activation it also goes the other "a! around: "henever mutual ad;ustments in the inferential steps occur then the "eight -i#e# the activation. of the related conte&tual items can be e&pected to change in turn# %n other "ords it is far from clear that associative processes can be restricted to a one directional flo" that stops "hen e&plicature and conte&tual assumptions are fed into the derivation# 0econd the assumption of a mutual ad;ustment ma(es the picture more comple& also in the sense that some sort of monitoring seems to be re4uired# %n a standard conception of inferential processes derivations can be thought of as blind applications of rules# 5ut in RT's frame"or( the mechanism needs to read < so to spea( < the changes intervening in subse4uent inferential steps in order to change accordingl! the previous steps of the derivation# An associative account has no difficulties in coping "ith this issue since changes in the activation of an! representation automaticall! affect the activation of others provided that there e&ist schemata associativel! bridging the t"o# On the other hand RT has never specified to m! (no"ledge ho" this monitoring and ad;usting "ould be accomplished: it is one thing to 6> sa! that a mutual ad;ustment bet"een premises and conclusions has to occur it is 4uite another to provide an e&plicit cognitive e&planation of ho" this can be done# %n this sense either the associative d!namic is allo"ed to e&tend be!ond the determination of premises and conclusions and to accomplish the assessment of the inferential structure as "ell or else RT needs to provide a different account of ho" the monitoring'and'ad;usting process is accomplished# Third it is in fact not clear "h! one should presume that the surrounding associative d!namics has to stop on the boundar! of the inferential component# Jet us recall for e&le the discussion of -?. and in particular the schema invo(ed b! =arston -6772. in order to arrive at the right interpretation of 0he missed her coach: -2. M%00%HG A @C0%GHATC@ =OA=$ 6 %0 A RCA0OH BOR A ,CR0OH HOT ARR%M%HG :$CH CN,C=TC@# Once "e admit that such a schema is part of the sub;ects' stored (no"ledge there is no reason to assume that it is not involved in the general associative d!namic of our mind: not onl! can the schema be activated b! conte&tual items it also can activate in turn other representations counting as -components of. premises or conclusions "ithin ;ustificator! structures# Recanati -677>. has proposed that schemata such as -2. add activation to an! item that happen to fit them# %t is not difficult to imagine that activation ma! similarl! propagate from one step to another of ;ustificator! structures -i#e# premises and conclusions. either because the! share some concepts or because there are abstract schemata connecting the premises and conclusions involved# Generall! spea(ing there is no reason to thin( that the associative d!namic has to stop at the point at "hich assumptions are fed to the alleged inferential component# These considerations seem to sho" that in accounting for mutual ad;ustment in pragmatic derivations it is not eas! to dra" a demarcation bet"een the associative stage and an alleged inferential process that "ould differ in (ind# 0pecificall! it is not clear "h! the associative d!namic should stop "hen premises and conclusions are set -first and third point. and if this "ere the case ho" then the monitoring'and'ad;usting step "ould be accomplished -second point.# RT seems to assume that "hile premises -and possibl! conclusions. of pragmatic inferences are accessed through a merel! associative route the! are evaluated b! a different cognitive process# =ontrar! to this h!pothesis m! previous considerations aim to sho" that the same d!namic of activation mediated b! schemata can e&plain both 68 ho" premises and conclusion-s. of pragmatic inferences are accessed and ho" premises and conclusion-s. are ad;usted to each other and to the conte&t in a motivated manner# %n sum RT is clearl! committed to a 4uite non'standard account of inferential processing in pragmatics according to "hich inferences are automatic and based on parallel mutual ad;ustment# This vie" does not seem prone to the continuum problem affecting the more standard Gricean accounts and it seems naturall! accommodated "ithin a thoroughl! associative vie"# $o"ever relevance theorists seem to insist instead on a cognitive discontinuit! bet"een the associative and the inferential component of the process# % have argued though that the thesis of a cognitive discontinuit! raises some problems# Birst there is no reason to assume that the associative d!namic has no resources to ensure the rational ad;ustment of premises and conclusion-s. to each other and to the conte&t# And second if this "ere the case RT "ould be in need to specif! "hich mechanism can ensure such ad;ustment# All this is not to sa! that associative processes are all there is to pragmatic processing# On the contrar! % have recentl! proposed that in language comprehension pure spreading activation must be complemented b! "or(ing memor! and attentional processes -Mazzone 67/9 b 67/9 c.# $o"ever although % cannot discuss the issue in detail here the distinction bet"een automatic and attentional processes can hardl! correspond to the distinction bet"een forming interpretive h!potheses and confirming+disconfirming them# %n the first place attentional processes can be thought to differ from automatic associative ones essentiall! in terms of the scale of activation+inhibition in the brain# As a matter of fact the general d!namic of activation and inhibition in the brain ta(es place at different scales spanning from most local to most global "ith -most. global processes implementing e&ecutive functions and specificall! attentional processing# Thus there is no discontinuit! in the strict sense bet"een associative and attentional processes: the! are part of the same d!namic of activation+inhibition ;ust at different scales# %n the second place attentional processes do not provide ne" information or specialized rules in order for rational decisions to be made# The! change instead the overall d!namic of activation from local to global < and from parallel to serial < so that onl! part of the information associativel! available to the s!stem for the purpose of rational assessment is specificall! focussed and assigned a role in activel! driving cognition# Therefore although in the normal case conscious attention is involved in utterance understanding it "ould not be correct to sa! that in principle attention is necessar! in order for interpretive h!potheses to 6? be confirmed+disconfirmed# On the contrar! automatic associative processes have in principle the resources to accomplish the mutual ad;ustment bet"een premises and conclusions in a motivated manner# (. &onclusions %n this paper % have focussed on a problem for Grice's account of le&ical pragmatics "hich % have referred to as the continuum problem and % have proposed a solution to it# The e&istence of a continuum bet"een le&icalised meanings and alleged implicatures raises a problem for both Modified Occam's Razor and the ver! notion of implicature at least as it is used in le&ical pragmatics# The argument according to "hich in case of uncertaint! bet"een conventionalisation and implicatures "e should prefer an account based on the latter loses its grip once "e recognize that there can be intermediate cases bet"een pure le&icalised meanings and pure implicature'li(e cases# Burthermore those intermediate cases raise a problem for a cognitive interpretation of the notion of implicature since "hile it is easil! agreed that conventionalisation ma! come in degrees it is far from clear that the processes delivering implicatures ma! decline accordingl!# A possible solution to this problem has been found in an associative account "here implicature'li(e effects are e&plained as the result of schemata "hich constrain the selection of conte&tuall! appropriate interpretations# Ac)no*ledgments A previous draft of this paper has been discussed in a seminar at Iniversit! =ollege Jondon in Ma! 67/9# % "ant to than( all the participants < 5arbara Cizaga Rebollar Alison $all Niaohong Giang @iana Mazzarella %sabelle Heedham'@idsbur! and especiall! Rob!n =arston and @eirdre :ilson < for their precious comments# % have also to than( the anon!mous referees for their insightful comments# 62 References 5ach )ent# /118# 0tandardization vs# =onventionalization# Linguistics and ,hilosophy /3: ?22' ?3?# 5ach )ent# /113# 0tandardization revisited# %n A# )asher -ed#. ,ragmatics/ 'ritical concepts0 *ol "*01 2/6'66# Jondon and He" Ror(: Routledge# 5arsalou Ja"rence# /116# 'ognitive ,sychology/ An (verview for 'ognitive Scientists# $illsdale H#G#: Ja"rence Crlbaum Associates# 5ontl! Thomas @# 6778# Modified Occam's Razor: ,arsimon! ,ragmatics and the Ac4uisition of :ord Meaning# %ind 2 Language 67 -9.: 633'9/6# =arston Rob!n# 6772# $o" man! pragmatic s!stems are thereA %n M#G# Brapolli -ed#. Saying1 %eaning1 #eferring0 Essays on the ,hilosophy of 3ran4ois #ecanati1 /'/2# He" Ror(: ,algrave# =arston Rob!n# 67/7# C&plicit =ommunication and 'Bree' ,ragmatic Cnrichment# %n 5#0oria and C# Romero -eds#. E.plicit 'ommunication/ #obyn 'arston5s ,ragmatics1 6/2'32# He" Ror(: ,algrave# =roft :illiam and @avid Alan =ruse# 677># 'ognitive Linguistics# =ambridge: =ambridge Iniversit! ,ress# @ehaene 0tanislas Gean ,ierre =hangeu& Jionel Haccache Gerome 0ac(ur and =laire 0ergent# 677?# =onscious preconscious and subliminal processing: a testable ta&onom!# Trends in 'ognitive Sciences /7-8.: 67><//# @evitt Michael# 677># The case for referential descriptions# %n M# Reimer S A# 5ezuidenhout -eds#. $escriptions and 6eyond1 637'978# O&ford: O&ford Iniversit! ,ress# @evitt Michael# 6772# Referential @escriptions and =onversational %mplicatures# European 7ournal of Analytic ,hilosophy 9/ 2'96# Gazdar Gerald# /121# ,ragmatics0 "mplicature1 ,resupposition1 and Logical 3orm# He" Ror(: Academic ,ress# Goldberg Adele# /118# 'onstructions/ A 'onstruction 8rammar Approach to Argument Structure# =hicago: Iniversit! of =hicago ,ress# Grice $erbert ,aul# /123# Burther Hotes on Jogic and =onversation# %n ,# =ole S G# J# Morgan -eds#. Synta. and Semantics1 vol0 91 ,ragmatics1 /39'12# He" Ror(: Academic ,ress# Grice $erbert ,aul# /131# Studies in the +ay of +ords# =ambridge MA: $arvard Iniversit! 63 ,ress# $orn Jaurence R# /126# (n the Semantic ,roperties of Logical (perators in English# 5loomington: %IJ=# $orn Jaurence R# /13># To"ard a ne" ta&onom! for pragmatic inference: K'based and R'based implicature# %n @# 0chiffrin -ed#. %eaning1 3orm1 and :se in 'onte.t1 //'>6# :ashington: Georgeto"n Iniversit! ,ress# $orn Jaurence R# and 0amuel 5a!er# /13># 0hort'circuited implicature: A negative contribution# Linguistics and ,hilosophy 2: 912'>/># Mazzocco Michele# /112# =hildren's interpretations of homon!ms: A developmental stud!# 7ournal of 'hild Language 6>: >>/'?2# Mazzone M# 67//: 0chemata and associative processes in pragmatics# 7ournal of ,ragmatics1 >9 6/>3<6/81# Mazzone M# 67/9 a: A pragmatic ,andora's bo&# Regularities and defaults in pragmatics# %n B# Jiedt(e and =# 0chulze -eds#. 6eyond +ords0 'ontent1 'onte.t1 and "nference# 5erlin: Mouton de Gru!ter# Mazzone M# 67/9 b# Attention to the spea(er# The conscious assessment of utterance interpretations in "or(ing memor!# Language 2 'ommunication 99-6. /7?'//># Mazzone M# 67/9 c: Automatic and controlled processes in pragmatics# %n A# =apone B# Jo ,iparo M# =arapezza -eds#. ,erspectives on Linguistic ,ragmatics# 5erlin: 0pringer# Mazzone M# in press =rossing the associative+inferential divide: ad hoc concepts and the inferential po"er of schemata Revie" of ,hilosoph! and ,s!cholog!# Mazzone M in preparation $o" concepts are formed# The relationship bet"een schemata theories and similarit! Morgan Gerr!# /123# T"o t!pes of convention in indirect speech acts# %n ,# =ole -ed#. Synta. and Semantics1 vol0 91 ,ragmatics1 6?/<637# He" Ror(: Academic ,ress# Hunberg Geoffre! @# /122# The pragmatics of reference# ,h@ dissertation =it! Iniversit! of He" Ror(# ,hillips 5en# 67/6# Modified Occam's Razor# Australasian 7ournal of ,hilosophy 17 -6.: 92/' 93/# 61 Recanati BranTois# 677># Literal %eaning# =ambridge: =ambridge Iniversit! ,ress# Recanati BranTois# 6772# Repl! to =arston# %n M# G# Brapolli -ed#. Saying1 %eaning1 #eferring0 Essays on the ,hilosophy of 3ran4ois #ecanati1 >1<8>0 He" Ror(: ,algrave# Reimer Marga# /113# @onnellanQs distinction+)rip(eQs test# Analysis 83: 31</77# Rubio Bernandez ,aula# 6772# 0uppression in metaphor interpretation: @ifferences bet"een meaning selection and meaning construction# 7ournal of Semantics1 6>: 9>8'2/# 0adoc( Gerrold# /123# On testing for conversational implicature# %n ,# =ole -ed#. Synta. and Semantics1 vol0 91 ,ragmatics1 63/<612# He" Ror(: Academic ,ress# 0earle Gohn# /128# %ndirect speech acts# %n ,# =ole and G# Morgan -eds#. Synta. and Semantics ;/ Speech Acts1 81'36# He" Ror(: Academic# 0ober Clliott# /133# #econstructing the ,ast# =ambridge MA: M%T ,ress# 0perber @an# 6778# Modularit! and relevance: $o" can a massivel! modular mind be fle&ible and conte&t'sensitiveA %n ,# =arruthers 0# Jaurence and 0# 0tich -eds#. The innate mind/ Structure and content1 89<?3# O&ford Cngland: O&ford Iniversit! ,ress# 0perber @# S :ilson @# -6773.# A deflationar! account of metaphors# -%n R# :# Gibbs -ed#. The 'ambridge handbook of metaphor and thought -pp#3>'/78.# =ambridge: =ambridge Iniversit! ,ress#. :al(er Ralph# /128# =onversational implicatures# %n 0# 5lac(burn -ed#. %eaning1 #eference1 and )ecessity1 /99<3/# =ambridge: =ambridge Iniversit! ,ress# :ilson @eirdre 677># Relevance and le&ical pragmatics# :'L +orking ,apers in Linguistics /?: 9>9'9?7# :ilson @eirdre and Rob!n =arston# 677?# Metaphor relevance and the emergent propert! issue# %ind and Language 6/ -9.: >7><>99# :ilson @eirdre and Rob!n =arston# 6772# A unitar! approach to le&ical pragmatics: relevance inference and ad hoc concepts# %n H# 5urton'Roberts -ed#. ,ragmatics1 697<6?7# He" Ror(: ,algrave# 97