Deshors & Gries, A Case for the Multifactorial Assessment of Learner Language. the Uses of May and...

8/18/2019 Deshors & Gries, A Case for the Multifactorial Assessment of Learner Language. the Uses of May and Can in Frenc…

1/27

Tis is a contribution from Corpus Methods for Semantics. Quantitative studies in polysemyand synonymy .Edited by Dylan Glynn and Justyna A. Robinson.© . John Benjamins Publishing Company

Tis electronic le may not be altered in any way.Te author(s) of this article is/are permitted to use this PDF le to generate printed copies tobe used by way of offprints, for their personal use only.Permission is granted by the publishers to post this le on a closed server which is accessibleto members (students and staff ) only of the author’s/s’ institute, it is not permitted to postthis PDF on the open internet.For any other use of this material prior written permission should be obtained from thepublishers or through the Copyright Clearance Center (for USA: www.copyright.com ).Please contact [email protected] or consult our website: www.benjamins.com

ables of Contents, abstracts and guidelines are available at www.benjamins.com

John Benjamins Publishing Company

http://www.copyright.com/mailto:[email protected]://www.benjamins.com/http://www.benjamins.com/http://www.benjamins.com/http://www.benjamins.com/mailto:[email protected]://www.copyright.com/


2/27

A case for the multifactorial assessmentof learner languageTe uses o may and can in French-Englishinterlanguage

Sandra C. Deshors and Ste an T. GriesNew Mexico State University / University o Cali ornia, Santa Barbara

In this study, we apply Gries and Divjak’s Behavioral Prole approach to com-pare native English can and may , learner English can and may , and French pou-voir . We annotated over 3,700 examples across three corpora according to morethan 20 morphosyntactic and semantic eatures and we analysed the eatures’distribution with a hierarchical cluster analysis and a logistic regression. Tecluster analysis shows that French English learners build up airly coherent cate-gories that group the English modals together ollowed by pouvoir , but that theyalso consider pouvoir to be semantically more similar to can than to may . Teregression strongly supports learners’ coherent categories; however, a variety ointeractions shows where learners’ modal use still deviates rom that o nativespeakers.

Keywords: Behavioral Proles, hierarchical cluster analysis, logistic regression,modal verbs

. Introduction and overview

Acquiring a oreign language is one o the most cognitively challenging tasks, giv-en how languages differ in every level o linguistic analysis. From a cognitively andpsycholinguistically-oriented perspective, learning a language requires identi ying a very large amount o co-occurrence data – tense t and number n require subject-verb

agreement with morpheme m, idiom i consists o word w and word x , communica-tive unction f is communicated with intonation curve c, etc. – as well as storing andretrieving them. Crucially, these types o co-occurrences are typically probabilisticonly rather than absolute/deterministic and, thus, hard to discern and learn: usually,learners need to cope with many-to-many mappings between orms and unctions,


3/27

Sandra C. Deshors and Ste an T. Gries

and ofen it is only the conuence o differently predictive in ormation on severallevels o linguistic analysis that narrows down the search or a particular meaning (in

comprehension) or a particular orm (in production). In the Competition Model byBates and MacWhinney (1982, 1989), or example, this situation is modeled on theassumption that orms and unctions are cues to unctions and orms, respectively,and many different cues o different strengths, validities, and reliabilities must be in-tegrated to, say in production, arrive at natural-sounding choices.

Semantics is a particularly tricky linguistic domain in this regard, in native lan-guage, but even much more so in oreign language learning. Not only do languagesofen carve up semantic space very differently (so that the categories o the languageacquired rst will inuence category ormation in the ollowing), but semantic differ-ences are also ofen much less explicitly noticeable (than, say, the presence or absenceo a plural morpheme), which makes the identication o probabilistic co-occurrencepatterns all the more difficult. In order to allow or a precise description o semantic,or more generally unctional, characteristics o synonyms, antonyms, and senses opolysemous words, Gries and Divjak developed the so-called Behavioral Prole (BP)approach (c . Gries and Divjak 2009). Tis approach, to be discussed in more detailbelow, is highly compatible with a psycholinguistic perspective o the type outlinedabove and involves a very ne-grained annotation o corpus data as well as their sta-tistical analysis.

Te method o behavioral proles has been success ully employed in a varietyo contexts – synonyms, antonyms, and word senses o polysemous words have beenstudied both within one L1 or across two different L1s – as well as having received rstexperimental support, but so ar there have been no studies that test the BP approach’sapplicability to L1 and L2 data, which is what we will undertake here. Te semanticdomain we will explore is one that has proven particularly elusive, namely, modality.While many semantic phenomena can be clearly delineated and, to some degree, ex-plained by the linguistic analyst, modality has been much more problematic; in act,

even the scope o the notion o modality has not really been agreed upon yet. In thischapter, we specically ocus on the semantic domain o as reected in:

– the choices o can vs. may in essays written by native speakers o English;– the choices o can vs. may in essays written by French learners o English;1

– the use o pouvoir in essays written by native speakers o French.

In Section 2, we discuss in what sense these modals pose a particular challenge to theanalyst as well as present previous corpus-based work on can and may and highlight

. Following Bartning (2009), the term “advanced learner” is hence orth assumed to re erto “a person whose second language is close to that o a native speaker, but whose non-nativeusage is perceivable in normal oral or written interaction” (Hyltenstam et al. 2005: 7, cited inBartning 2009: 12).


4/27

A case or the multi actorial assessment o learner language

some o the shortcomings o such work. In Section 3, we discuss the BP approachin general as well as our own data and methods in particular. Section 4 presents the

results o our exploration, and Section 5 concludes the chapter.

. Setting the stage

. What is problematic about the modals?

As near synonyms in the domain o modality, may and can have ueled much the-oretical debate with regard to their semantic relations. As a pair, both orms haveoverlapping semantics which cover simultaneously the meanings o possibility, per-mission and ability (c . Collins 2009). Tis means that both orms can be used to ex-press epistemic, deontic and dynamic types o possibility. It ollows that the semanticinvestigation o may and can triggers two problematic questions: rst, to what extentthe various senses o each orm can be distinguished, and second, to what degree both

orms are semantically equivalent?With regard to the rst question, studies such as Leech (1969) and Coates (1983)

have illustrated the difficulty in distinguishing between the senses o may and can.

Leech (1969: 76), or instance, notes that “[t]he permission and possibility meaningso may are close enough or the distinction to be blurred in some cases”. Similarly, Coates (1983: 14) identies a “continuum o meaning” – i.e. gradience – in whichpossible modal uses shade into each other. In the case o the meanings o can, orinstance, Coates notes that while permission and ability correspond to the core o twolargely intersecting uzzy semantic sets, possibility, on the other hand, is ound “in theoverlapping peripheral area” (p. 86).

With regard to the issue o the semantic equivalence o may and can, the literaturereveals similarly debated standpoints. While some studies recognize the similarities

o the two orms, others do not. In the ormer case, or instance, Collins (2009: 91)states that “[t]he two modals o possibility may and can, share a high level o seman-tic overlap” (despite their differing requency o occurrence and different degrees o

ormality), and Leech (1969: 75) notes that “[i]n asking and giving permission, canand may are almost interchangeable”. Conversely, studies such as Coates (1983) haveclearly distinguished the two orms. For instance, while Coates (1983) does recognizethat the English modals share certain meanings and can be organized into semanticclusters, she generally denies the synonymy o may and can by classi ying the two

orms into two distinct semantic groups. Although she accepts that the two ormsmay have overlapping meanings in some cases, she claims that even then, the two

orms do not occur in ree variation.Te occurrence o one orm over the other has been shown to be inuenced, to

some extent, by its linguistic context. It has indeed been illustrated that particular


5/27


co-occurring grammatical categories inter ere with the interpretation o the modals.Leech (2004: 77), or instance, notes that certain uses o may are only to be ound in

particular grammatical contexts: “only the permission sense, or instance, is oundin questions (…) and the negation o the possibility sense is different in kind romthe negation o the permission sense”. Generally, several grammatical categories havebeen recognized as interacting with the uses o may and can. While negation is onecategory that has commonly been identied (c . Hermerén 1978; Palmer 1979; Coates1980, 1983; De Haan 1997; Huddleston 2002; Radden 2007; Byloo 2009), voice andsentence types have also been shown to have similar inuences on the orms.

Overall, the above-mentioned studies all provide clear illustrations o the com-plexity o the semantic relations between may and can on the basis o empiricallygathered evidence. However, they all tend to be based on generalized observations oidiosyncratic behavioral tendencies. In that respect, they all raise the issue o how toprovide a more systematic account o the modals’ semantic characteristics and how tointegrate qualitative ndings into a quantitative and empirically-grounded approach.

. Previous corpus-based work on the modals

. . Native EnglishAs already mentioned above, Hermerén (1978) has shown that the semantics o themodals in native English are morphosyntactically motivated to a considerable de-gree such that linguistic categories such as voice, grammatical person, type o main verb (action, state, etc.), aspect and sentence type inuence the interpretation o themodals: “i these categories can be shown to modi y the meaning o the modal […] itis important that this should be accounted or in the description o the semantics othe modals” (p. 74). While this claim calls or empirical validation, one implication oHermerén’s (1978) argument is that the quantitative study o modal orms will requirea power ul and versatile methodological approach. In a very similar ashion, Klingeand Müller (2005: 1) argue that, to capture the essence o modal meaning, “it seemsnecessary to cut across the boundaries o morphology, syntax, semantics and prag-matics and all dimensions rom cognition to communication are involved”.

A second corpus-based study o the modals in native English is Gabrielatos andSarmento (2006). Tis study illustrates an attempt to account or syntactic contextu-al in ormation while using a quantitative corpus-based approach to investigate coreEnglish modals (i.e. can, could , may , might , must , shall , should , will and would ). Al-though their study does not involve the comparison o English varieties, it presents,

however, a comparative analysis o the requencies o uses o the modals in an aviationcorpus and a representative corpus o American English. Generally, it raises the ol-lowing questions:


6/27


– o what degree do syntactic structures and modal orms interact contextually?– o what degree does such interaction affect investigated modal orms semantically?

– How can such interaction be quantitatively investigated in a corpus includingcross-linguistic and interlanguage data?

Te authors acknowledge that the modals’ distribution varies as a unction o theirsyntactic contexts and they show that requencies o occurrence o core Englishmodals reect the type o syntactic environment in which they eature: “there is agreat deal o variation in the use o modal verbs and the structures they occur in,depending on the context o use” (p. 234). However, their lack o a suitable cogni-tively-motivated theoretical ramework prevents them rom providing a meaning ul

interpretation o the data and to urther explore their ndings.o this date, Collins (2009: 1) presents:

the largest and most comprehensive [study] yet attempted in this area [modality]based on an analysis o every token o the modals and quasi-modals (a total o46,121) across the spoken and written data.

Collins (2009) investigates the meanings o the modals in three parallel corpora ocontemporary British English, American English and Australian English. Despite theauthor’s recognition that a corpus quantitative approach “typically combined with acommitment to the notion o ‘total accountability’ may inuence hypotheses appliedto the data, or ormulated on the basis o it” (p. 5) and despite the large size o his dataset, his analysis is o limited in ormative value due to:

– a theoretical ramework that does not allow or the ull exploitation o the linguis-tic context o the modals, and;

– a statistical approach that inhibits rather than unveils linguistic patterns at play inthe data.

With regard to the rst point, Collins (2009) restricts his approach to the identica-tion o the orms’ lexical meanings. His theoretical ramework consists o a tradition-al tripartite taxonomy including epistemic, deontic and dynamic senses. Regrettably,while he recognizes that some uses o the modals can yield pre erences or particularsyntactic environments, his analysis does not address that act in a systematic quan-titative ashion. As or the second point, while, statistically, Collins (2009) limits hisinvestigation to providing requency tables o modal orms, his overall approach isproblematic because it is based on the erroneous assumption that the requent occur-rence o a modal orm warrants its linguistic relevance. In the case o may and can, or

instance, Collins uses raw requencies to show that deontic may is the “least common”sense o the three as it is chosen 7% o the time over epistemicmay (79%) and dynam-ic may (8.1%). However, he does not show whether the (low) requency o deonticmay is signicantly different rom the also low requency o dynamicmay , and our


7/27


analysis o his data shows that, excluding the indeterminate cases, the distribution omay’ s senses across the American, Australian, and British data is highly signicant

( χ 2 = 42.68; df = 4; p < 0.001). Tis, in turn, raises the questions o :

– o what extent are Collins’ (2009) requencies o the occurrences o modal ormsin each corpus comparable?

– Since the observed requency discrepancies are not a matter o chance, then whatmotivates, linguistically, the different uses o each orm in each independentcorpus?

So in sum, while studies such as Gabrielatos and Sarmento (2006) and Collins (2009)provide many descriptive results, they are ofen merely or largely orm-based aloneand are lacking in terms o determining which o the many requencies are statistical-ly and/or linguistically relevant. As a result, such studies do not come close to allowus to develop a characterization o modals that essentially allows us to classi y/predictmodal use.

. . Learner English and contrastive approachesFrom a cross-linguistic and an interlanguage perspective, investigating the modalsraises two related issues, namely (i) the possibility o a lack o (direct) semantic equiv-

alence between the modal orms in the learner’s native language (L1) and his/hertarget language (L2), and (ii), the act that such cross-linguistic semantic dissimilaritywill affect the uses o the orms in L2. Te modals may and can and native French pouvoir illustrate the case in point. Despite the act that all three orms contribute tothe expression o the semantic notion o , pouvoir synchronically coversthe whole range o the modal uses o may and can.

One corpus-based study o learners’ use o modals is Aijmer (2002), which isbased on a corpus o Swedish L2 English writers. She compares (i) the requencieso key modal words in native English and advanced Swedish-English interlanguage,as well as (ii) requencies encountered in Swedish learner English with those romcomparable French and German L2 English. Aijmer’s study indicates “a generalizedoveruse o all the ormal categories o modality” and she urther points out that “itis only at a unctional level that any underuse was detected, with the learner writers

ailing to usemay at all in its root meaning” (p. 72).Similarly, Neff et al. (2003) investigate the uses o modal verbs (can, could , may,

might and could ) by writers rom several L1 backgrounds. Neff et al. (2003) use alearner corpus including Dutch-, French-, German-, Italian-, and Spanish-Englishinterlanguage, which they contrast with a re erence corpus o American universityEnglish. Neff et al. (2003: 215) identi y the case o can as potentially interesting “sinceit is overused by all non-native writers”. Tey urther report that the requency o may by French native speakers stands out in comparison to the requencies by all othernon-native speakers included in the study, but since their study does basically nothing


8/27


but compare raw requencies o occurrence regardless o any contextual eatures , it isnot particularly illuminating.

Generally, and similar to Gabrielatos and Sarmento (2006) and Collins (2009),both Aijmer (2002) and Neff et al. (2003) made the disadvantageous methodologicaldecision to conveniently, but ultimately problematically, rely on in ormation that isretrievable without human effort. In addition, even the studies that address learneruse do not relate their ndings to the wider context o (second) language acquisition.

In a corpus-based contrastive study, Salkie (2004) investigates the nature o thesemantic relations between the three orms in native English and native French. Heuses a subpart o the parallel corpus IN ERSEC (c . Salkie 2000), and ocuses onthree working hypotheses, namely that:

– “ pouvoir corresponds more closely to one o the English modals rather than theother” (p. 169);

– “ pouvoir is less specic than the English modals” (p. 170);– “ pouvoir has a sense which is different rom both the English modals but is not

just a general sense o possibility” (p. 170).

While Salkie (2004) concludes in avour o the third hypothesis, it is worth point-ing out, however, that his results were based on only 100 randomly extracted occur-

rences o each English modal orm (i.e. may and can) and their respective Frenchtranslations.By way o a more general summary, it is probably air to say that corpus-based

approaches to modality in L1 and L2s leave things to be desired. Some studies pointto the immense complexity o the subject but do not choose multi actorial or multivariate methods that are capable o addressing this degree o complexity. In addition,some studies are based on large numbers o modals but, rankly, do not do very muchwith the vast amount o data other than present arrays o statistically under-analyzed

requency tables. On the other hand, the analytically much more interesting studies

o the kind o Salkie (2004) are based on very small samples. Finally, many studies arelargely i not exclusively orm-based and ocus only on learners’ over-/underuse omodals in particular examples or kinds o contexts.

. Characteristics o the present study

. . Methodological considerationsTe above discussion airly clearly indicates what kinds o steps would be desirable,

an approach that:– can integrate linguistic in ormation and patterning rom many different levels o

linguistic analysis in a way alluded to by Hermerén (1978), as well as Klinge andMüller (2005);


9/27


– involves not only a sample that is studied with regard to more linguistic parame-ters, but at the same time also larger than the previous studies that aimed at more

than description;– explores similarities and differences o L1 uses o can and may , but also explores

the way these English modals are used in L2 language (here rom French learners)as well as how the same concept is used by the learners in their L1 (here pouvoir ).

Given these demands, we decided to use the so-called Behavioral Prole approach,which ts the above wish list very well. It combines the statistical methods o con-temporary quantitative corpus linguistics with a cognitive-linguistic and psycholin-guistic perspective or orientation (c . Divjak and Gries 2006, 2008, 2009; Gries 2006,

2010b; Gries and Divjak 2009, 2010; and others). As such, it diverges radically romthe above-mentioned more traditional corpus-based approaches to modality in bothL1 and L2. Methodologically, it involves our steps:

– the retrieval o all instances o a word’s lemma rom a corpus in their context;– a manual annotation o a number o eatures characteristic o the use o the word

orms in the data; these eatures are re erred to as ID tags and typically involvemorphosyntactic and semantic eatures in particular. Each ID tag contributes tothe proling o the investigated lexical item(s);

– the generation o a table o co-occurrence percentages, which speci y, or exam-ple, which words ( rom a set o near-synonymous words) or senses (o a polyse-mous word) co-occur with which morphosyntactic and/or semantic ID tags; it isthese vectors o percentages that are called proles;

– the evaluation o that table by means o statistical techniques.

Given how this approach is completely based on various kinds o co-occurrence in-ormation, it comes as no surprise that, just like much other work in corpus linguis-

tics, the BP approach assumes that “the distributional characteristics o the use o an

item reveals many o its semantic and unctional properties and purposes” (Gries andOtani 2010: 3). While these previous studies have investigated a variety o differentlexical relations (near synonymy, polysemy, antonymy) both within languages (Eng-lish, Finnish, Russian) and across languages (English and Russian), the present studywill add to the domains in which Behavioral Proles have been used in two ways:(i) so ar, no non-native language data have been studied, and (ii) we will add Frenchto the list o languages studied.

As the rst BP study ocusing on learner data, and only the second BP study thatcompares data rom different languages, this paper is still largely exploratory. We will

mainly be concerned with the ollowing two issues:

– o what degree can the Behavioral Proling handle the kind o learner data thatare inherently more messy and volatile than native data and provide a quantita-tively adequate and ne-grained characterization o the use o can and may by


10/27


native speakers and learners, and how does that use compare to the use o Frenchspeakers’ use o pouvoir ?

– As a ollow-up, and i meaning ul groups o uses emerge, to what degree do thedistributional characteristics that BP studies typically include allow us to predictnative speakers’ and learners’ choices o modal verbs, and how do these speakergroups differ?

Te ormer question will be explored with the kind o cluster-analytic approach usu-ally employed in BP studies; or the latter question, we will turn to a logistic regression(c . Arppe 2008 or another BP approach using (multinomial) regression).

. . Teoretical orientationIn previous studies, the BP approach was used or more than just the quantitativedescription o the data. Rather, it is rmly grounded in, and attempts to relate theresults o the statistical exploration o the data to usage-based/exemplar-based ap-proaches within Cognitive Linguistics and psycholinguistics. While this orientationis also compatible with our current goals, there is one particular earlier model in L2/FLA research that is especially well-suited to, or compatible with, our current objec-tives, namely the Competition Model (CM) by Bates and MacWhinney (c . Bates andMacWhinney 1982, 1989). Tis model is “a probabilistic theory o grammatical pro-cessing which developed out o a large body o crosslinguistic work in adult and childlanguage, as well as in aphasia” (Kilborn and Ito 1989: 261). MacWhinney (2004: 3)himsel characterized it as a “unied model [o language acquisition] in which themechanisms o L1 learning are seen as a subset o the mechanisms o L2 learning”.

Te CM is characterized by the two ollowing assumptions:

– Linguistic signs map orms and unctions onto each other (probabilistically) suchthat orms and unctions are cues to unctions and orms respectively.

– In language production, orms compete to express underlying intentions or unc-tions, and in language comprehension, the input contains many different cues odifferent strengths, validities, and reliabilities, which must be integrated: nativespeakers “depend on a particular set o probabilistic cues to assign ormal sur acedevices in their language to a specic set o underlying unctions” (Bates andMacWhinney 1989: 257).

As a usage-based and probabilistic model, the CM assumes that both requency andunction determine the choice o grammatical orms in language production; as with

most usage-based and/or corpus-linguistic approaches, we too consider requency in

a corpus as a proxy or requency o exposure (in both comprehension and produc-tion). Cross-linguistically, this is an important assumption because across languagescues are instantiated in different ways and speakers assign them varying degrees ostrength. It is there ore important to describe and explain L1 statistical regularities as


11/27


“[t]hey are part o the native speaker’s knowledge o his/her language, and they are animportant source o in ormation or the language learner” (Bates and MacWhinney

1989: 15).Overall, Kilborn and Ito (1989: 289) conclude that existing psycholinguistic stud-

ies have success ully demonstrated that the CM is appropriate or the characterizationo learner language through cue distributions and they report “extensive evidence orthe invasion o L1 strategies into L2 processing”. In addition, it is also obvious howmuch the CM is compatible with a BP approach. Te main notions that drive theCompetition Model are cue strengths, validities, and reliabilities, and all o these areessentially conditional probabilities, i.e. percentages. While the BP approach as suchdoes not cover the ull complexity o how conditional cue strengths, validities, andreliabilities can interact, it is a use ul and experimentally validated (c . Divjak andGries 2008) approach employing a similar logic.

A theory o language trans er requires that we have some ability to predict wherethe phenomena in question will and will not occur. In this regard contrastiveanalysis alone alls short; it is simply not predictive. (Gass 1996: 324)

. Data and methods

. Retrieval and annotation

Te data are rom three untagged corpora: the French subsection o the InternationalCorpus of Learner English (hence orth ICLE-FR), the Louvain Corpus of Native EnglishEssays (LOCNESS), and the Corpus de Dissertations Françaises (CODIF). All corporaincluded in the present work were collected by the Centre or English Corpus Lin-guistics (CECL) at the Université Catholique de Louvain (UCL) and made availableto us by the Director o the Centre, Pro essor Sylviane Granger. ICLE-FR has a totalo 228,081 words, including 177,963 words o argumentative texts and 50,118 wordso literary texts. LOCNESS is a 324,304-word corpus that includes three sub-datasets: a 60,209-word-sub-corpus o British A-Level essays, a 95,695-word sub-corpuso British university essays and a sub-corpus o American university essays that has168,400 words. Te CODIF is a corpus o essays written by French-speaking un-dergraduate students in Romance languages at the Université Catholique de Lou- vain (UCL). CODIF also includes argumentative and literary texts and has a total o100,000 words.2

. In ormation on the total number o words eaturing in each individual text type (i.e. argu-mentative, literary) is not available.


12/27


Given the corpora’s compositions, the three corpora included in our study arehighly comparable. Tey all consist o written data produced by university students(ICLE, CODIF, the LOCNESS British and American university sections) or by stu-dents approaching university entrance (i.e. the LOCNESS British A-Level section). 3 All participants’ contributions are in the orm o an essay o approximately 500 wordslong. In terms o content, all essays deal with similar topics such as: crime, education,the Gul War, Europe, or university degrees.

Te data we subjected to the BP approach consist o instances o may and can innative English and French-English interlanguage as well as pouvoir in native French

rom the above corpora. Using scripts written in R (c . R Development Core eam2010), we retrieved 3,710 occurrences o the investigated modal orms rom allsub-corpora, which were imported into a spreadsheet sofware and annotated or 22morphosyntactic and semantic variables. 4 able 1 exemplies this database with a very small excerpt o these data, and able 2 presents the total range o variables in-cluded in the study and their respective levels.

For each variable, an encoding taxonomy was designed prior to annotation. Dueto the large number o variables included in this study and the absence o a number othem rom previous studies on the English modals, not all encoding taxonomies weretheoretically motivated. In cases where the annotation is not based on accounts romthe existing literature, a bottom-up approach was adopted or the identication o re-current eatures in the data. Tis procedure, or instance, was carried out in the case othe variable V S where, prior to annotation, recurrent semantic eatureswere identied as characteristic o the lexical verbs used alongside the modals.

. Te inclusion o the LOCNESS British A-Level section alongside sub-corpora solely in-

cluding university participants is not judged problematic as LOCNESS only involves Englishnative speakers whose level o English is not expected to develop any urther.

. Although the annotation process included a variable encoding the semantic role o thesubject re erent o the modals, this study does not account or that variable due to its high cor-relation with VOICE.

able 1. Excerpt o an annotation table including selected variables

C M C C U V S N R A

5 may native coordinate process ment/cog/emotional affirmative animate133 may native main state copula affirmative inanimate1760 may native main process ment/cog/emotional negative animate1886 can il coordinate process ment/cog/emotional affirmative animate2876 cannot il subordinate state abstract negative inanimate3540 peut r main process ment/cog/emotional negative animate3645 peuvent r subordinate process abstract negative inanimate


13/27


Because o space restrictions, we are not able to provide a more comprehensiveaccount o the annotation process (but c . Deshors 2010 or details). However, three variables – S , V , and V S – require some brie explanatorycomments.

. . Te variable SAs or S , the semantic category o modality includes a wide range o hetero-geneous meanings that many scholars have attempted to unite under a variety ocategorization systems (c . Palmer 1979; Coates 1983; Bybee and Fleischman 1995;Huddleston 2002; Nuyts 2006; Byloo 2009). While Depraetere and Reed (2006: 277)note that “in classi ying modal meanings, it is possible to use various parameters ascriterial to their classication”, this study assumes a coding taxonomy based on atraditional tripartite distinction between epistemic, deontic and dynamic meanings.

able 2. Overview o the variables used in the study and their respective levels

ype Variable Levels

data C native, interlanguage, FrenchG A (acceptability) yes, no

syntactic N (negation) affirmative, negatedS (sentence type) declarative, interrogativeC (clause type) main, coordinate, subordinate

morphological F can, may , pouvoir (and negated orms)S M : subjectmorphology

adj., adv., common noun, proper noun,relative pronoun, date, noun phrase, etc.

S P : subject person 1, 2, 3S N : subject number singular, pluralV active, passiveA per ect, per ective, progressiveM indicative, subjunctiveS R N : subjectre erent number

singular, plural

semantic S epistemic, deontic, dynamicS P weak, medium, strongU accomplishment, achievement, process,

stateV S abstract, general action, action incurring

trans ormation, action incurring move-ment, perception, etc.

R A : subject re erentanimacy

animate, inanimate

A : subject re erentanimacy type

animate, oral, object, place/time, mental/emotional, etc.


14/27


Following Nuyts (2006: 6), epistemic senses concern “an indication o the epistemicestimation, typically, but not necessarily, by the speaker, o the chances that the state

o affairs expressed in the clause applies in the world”. Consider (1) as an illustrationo epistemicmay :

(1) indeed, Europe 92 may lead to the disappearance o cultural differences

Following Palmer (1979: 58), deontic modality re ers to cases where “[b]y uttering amodal, a speaker may actually give permission ( may , can)”. (2) illustrates deontic can:

(2) i all public schools started to say you can only come here i you are Hispanicor i you are Polish, our schooling system would be in great chaos

Finally, dynamic meanings denote “an ascription o a capacity to the subject-partic-ipant o the clause (the subject is able to per orm the action expressed by the main verb in the clause)” (Nuyts 2006: 3). Generally, dynamic modality expresses the po-tentiality o an event occurring. Nuyt’s type o dynamic modality includes ability/ capability cases where the possibility o event occurrence stems rom the ability o the(grammatical) subject to carry out the event. In that regard, the term ability is not re-stricted to a ‘physical’ interpretation and equally applies to mental and technical typeso ability. Example (3) illustrates dynamic can:

(3) Mrs Ramsay is the central character because she can see the whole personalityo the other ones

Generally, our requencies o use o may and can in their different senses match thosepreviously encountered in existing studies solely concerned with the native use othe modals, such as Coates (1980) and Collins (2009). While Coates (1980: 218), orinstance, reports that “by ar the most common usage o may is to express epistemicpossibility”, she stresses the distinctive nature o the uses o may and can:

Te patterns resulting rom my analysis o the data (…) leads me to concludethat in normal everyday usage may and can express distinct meanings: may isprimarily used to express epistemic possibility, while can primarily expresses rootpossibility. 5

. . Te variable VTe variable V targets the lexical verbs with which the orms are used andcharacterizes their telicity. Conceptually, the variable V ollows Vendler(1967) in its recognition that the notion o time is crucially related to the use o a

. Coates (1980, 1983) categorizes modal meaning according to a two-way distinction thatincludes epistemic and non-epistemic modality. She re ers to the latter type as “root” modality.


15/27


verb and is “at least important enough to warrant separate treatment” (p. 143). Tis variable assesses:

– whether may and can have pre erences or lexical verbs denoting astate, a process,an accomplishment or an achievement ,6 and i so,

– it identies in which type o corpus pre erential patterns occur.

. . Te variable V SSimilarly to the variable V , V S identies the type o semanticin ormation conveyed by the lexical verbs used with the modals. Te internal organ-ization o this variable results rom a bottom-up approach and does not ollow any

particular theoretical ramework. Tis variable consists o the levels denoting abstractprocess, physical actions, actions incurring movement, actions incurring some physi-cal trans ormation, communicative processes, mental/cognitive/emotional processes,perception processes and verbal statement involving a copula verb. Example (4) illus-trates a case where the lexical verb expresses a mental/cognitive/emotional process:

(4) Her search or the nal touch can be seen as a search or harmony

Once all matches were annotated, the resulting data table was evaluated statistically.

. Te BP approach in this study: Statistical analysis

As mentioned above, the data were evaluated in two different ways. 7 Te rst o theseinvolved the type o cluster analysis that is characteristic o much work using the BPmethodology. In this rst part, we used Gries ’s (2010a) R script Behavioral Proles1.01 and computed ve behavioral proles, one or each modal orm as occurring ineach language variety, i.e. native can, native may , interlanguage (IL) can, IL may , andnative pouvoir (FR). Such proles consist o vectors o co-occurrence percentages o asingle modal orm with each level o all independent variables and provide orm-spe-cic summaries o their semantic and morphosyntactic behavior in each sub-cor-pus. In a second step, the proles were assessed statistically with a hierarchical clusteranalysis to explore the similarity and differences between the modal orms, and inkeeping with previous studies (c . Divjak and Gries 2006), we chose the Canberrametric as a measure o (dis)similarity and Ward’s rule as an amalgamation strategy.

. Accomplishment verbs encode verbal statements that imply a unique and denite time

period; achievement verbs encode verbal statements that imply a unique and denite time in-stant; process verbs identi y statements that reect non-unique and indenite time periods;state verbs identi y statements that reect non-unique and indenite time instants.

. All statistical computations and plots were per ormed with R ( or Linux), version 2.11.0(see R Development Core eam 2010).


16/27


Following Gries and Otani (2010), we computed different cluster analyses, one involving all variables that the uses o the modals were annotated or, one or only the

syntactic variables, and one or only the semantic variables.Te second analytical step involved a binary logistic regression including the ol-

lowing variables and predictors:

– F as the dependent variable with only two levels here: can vs. may ;– G A , N , S , C , S M , S P , S N ,

V , A , M , S R N , S , S P , U , V -S , R A , A as independent variables in the orm o maineffects;

– all these variables’ interactions with C as additional predictors (to seewhich variables’ inuence on modal use differs the most between L1 English andL2 English).

Te logistic regression was then per ormed with the model selection process duringwhich insignicant predictors were discarded rom the model: rst insignicant in-teractions, then individual variables that were not signicant and did not participatein a signicant interaction.

. Results and discussion

. Cluster analysis

Our rst cluster analysis yielded the results shown in Figure 1. Te lef plot is a den-drogram o the ve modal orms that were clustered; the right plot represents averagesilhouette widths or assuming two, three, and our clusters. Te average silhouettewidths point to a two-cluster solution, maybe a three-cluster solution, but the differ-ence is minor since the ormer would result in a French-vs.-English clustering, andthe latter in a French-vs.- can-vs.-may clustering. Tis is compatible with Salkie’s anal-ysis, who argued that pouvoir is very different rom bothcan and may , and intuitively,both these solutions “make sense”, which provides rst evidence in avor o the ap-proach. o anticipate the potential objection that this may seem trivial, let us mentionthat it is in act not. Te data in Figure 1 show that the BP vectors are good and robustdescriptors o how the modals behave because many other theoretically possible clus-ter solutions, such as the ones listed in (5), would not have made linguistic sense at all.

(5) a. {{{canil may native pouvoir }cannative}may il} b. {{{cannative may il pouvoir }canil}may native} c. {{canil may native} { pouvoir may il}cannative}


17/27


However, in what ollows we show that a ne-grained comparative description ocross-linguistic language varieties can be obtained by ocusing on differences betweenthe independent variables used or clustering. Consider Figure 2, which shows thedendrograms or all morphosyntactic variables and all the semantic variables in thelef and right panel, respectively.

Interestingly, the results show that the intuitively very reasonable dendrogram inFigure 1 is not replicated by looking at morphosyntax or semantics alone, which tosome extent at least contrasts with Gries and Otani’s results, where the results did notdiffer very much between the three clusterings. Te reasonable similarities o Figure 1emerge only when all variables are combined. In particular, in both panels o Figure 2canil and cannative are grouped together, but then the remaining orms are grouped di -

erently. In the morphosyntactic dendrogram, the two kinds o may are successivelyamalgamated and the French pouvoir is only added afer all English orms have been

7 0 0

7 5 0

8 0 0

8 5 0

P O U V O I R

. f r

C A N

. i l

C A N

. n a t i v e

M A Y

. i l

M A Y

. n a t i v e

9 0 0

9 5 0

H e

i g h t

0 . 0

0 1 2 3 4

0 . 2

0 . 4

0 . 6

0 . 8

1 . 0

( A v e r a g e

) s i

l h o u e

t t e w i

d t h s

Number of clusters in the solution

0.1 0.07

0.03

Figure 1. Dendrogram or all independent variables (il = interlanguage)

1 . 5

1 . 0

0 . 5

2 . 0

2 . 5

3 . 0

3 . 5

P O U V O I R

. f r

C A N

. i l

C A N

. n a t i v e

M A Y

. i l

M A Y

. i l

M A Y

. n a t i v e

M A Y

. n a t i v e

4 . 0

4 . 5

H e i g h t

8

6

1 0

1 2

1 4

P O U V O I R

. f r

C A N

. i l

C A N

. n a t i v e

1 6

1 8

Figure 2. Dendrograms or all morphosyntactic variables (lef panel) and all semantic variables (right)


18/27


clustered. In other words, morphosyntactically, we nd a clear English-French divide,but interlanguage may is too different rom native may to be grouped together. oidenti y the source o this difference, we used what in BP approaches has been calleda snakeplot, namely a plot o the pairwise differences between the percentages or, inthis case, may il and may native (c . Divjak and Gries 2009 or Gries and Otani 2010 ormore examples).

As indicated in Figure 3, the main morphosyntactic ways in which learners deviate rom native speakers are that learners underuse may in subordinate clauses andin negated clauses. Tis is in act an interesting nding because it means that learnersdispre er the rarer o the two modals – may – in those contexts which are alreadymorphosyntactically more challenging, as i using can is the de ault they resort towhen they are already under a higher processing load (c . the so-called complexity

principle).In the semantic dendrogram, by contrast, we nd a different patterning. Seman-

tically, canil and cannative are again very similar and grouped together early, but thenthe next clustering step groups the two orms o may together. However, interestingly,it is not the English orms that are then all grouped together – rather, contrary toSalkie’s earlier analysis, pouvoir is semantically more similar to can than may is.

. Logistic regression

Te model selection process involved thirteen steps during which insignicant predic-tors were discarded. Te nal and minimally adequate model includes 16 signicant variables and 6 signicant interactions and returned a highly signicant correlation:loglikelihood chi-square = 3296.47; df = 60; p < 0.001; the correlation between the

0 . 4

s u b o r d c l a u s e s

n e g a t e

d a f

f i r m a t

i v e

m a i n c l a u s e s

s e v e

r a l t i n

y o r n

o d i f f e

r e n c e

s

S y n

t a c t i c

d i ff e r e n c e s m a

t i n t e r l a n g -

m a y n a

t i v e

0 . 2

0 . 0

– 0

. 2

– 0 . 4

Figure 3. Snakeplot or most extreme differences between syntactic ID tags o may


19/27


observed orms – may vs. can – and predicted probabilities is very high: R2 = 0.955.Correspondingly, the model’s classicatory power was ound to be very power ul witha classication accuracy o 99%. able 3 summarizes all the signicant variables andinteractions yielded in the nal model.

Overall, the nal model includes one signicant interaction involving a morpho-

logical variable (out o seven morphological variables), two signicant interactionsinvolving syntactic variables (out o three syntactic variables) and three signicant in-teractions involving semantic variables (out o eight semantic variables). But what dothe interactions reect? Let us begin with C :C , as represented in Figure 4.

Te requencies o may and can differ with regard to the type o clauses in whichthey occur in native and learner English. Te (weak!) effect is that, in interlanguage

able 3. Overview o the results o the nal GLM model

Predictor Chi-square ( df ) Predictor Chi-square ( df )

C 24.9 (1) *** A 98.2 (11) ***G A 13.8 (1) *** V 55.0 (1) ***U 67.9 (1) *** S 47.2 (1) ***E 100.0 (2) *** N 87.2 (1) ***C 10.9 (1) *** S P 29905.9 (2) ***V 97.4 (2) *** C :C 60.0 (2) ***V S 384.9 (6) *** C :V S 32.2 (6) ***S P 26.6 (2) *** C :S N 37.4 (1) ***S N 1.3 (1) ns C :R A 122.2 (1) ***S M 49.1 (4) *** C :A 118.2 (11) ***R A 59.2 (1) *** C :N 12.0 (1) ***

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Coordinate Main

Interlanguage

Subordinate

1 . 0

can

may

can

may

can

may 1

. 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Coordinate Main

Native

Subordinate

1 . 0

can

may

can

may

can

may

Figure 4. Bar plots o relative requencies o C :C


20/27


English, can is more strongly pre erred over may in main clauses than it is in nativeEnglish.

While, as previously noted, existing literature concerned with the native use othe modals commonly recognizes negation as “an important aspect o modal mean-ing” (Hermerén 1978), our study not only conrms the need to include negation inan investigation o the uses o the modals but urther recognizes its signicance as amorphological criteria to assess interlanguage (dis)similarity. Consider Figure 5 orthe interaction C :N .

Figure 5 shows that, while all speakers pre er to use can in negated clauses, the in-terlanguage speakers do so more strongly. Tis result does not come as a surprise: Onthe one hand, this is also compatible with the complexity principle – negated clausesare more complex and pre erred with the more requent modal. On the other hand,where epistemic may not would be used in English, French speakers would tend touse a lexical verb along with the adverb peut-être to indicate the speaker’s uncertainty,as illustrated in (6):

(6) a. Tis may not be the case b. Ce n’est peut-être pas le cas

Consider Figure 6 or the interaction C :S N .While native speakers use can more ofen with singular subjects than with plural

subjects, it is the other way round with the learners, again a result compatible with thecomplexity principle.

While the native speakers’ choices o may and can do not vary much betweenanimate and inanimate subjects, the learners’ choices do: with animate subjects, theypre ercan much more strongly. Figure 7 represents the interaction C :R A .

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Affirmative

Interlanguage

Negative

can

may

can

may 1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Affirmative

Native

Negative

can

may

can

may

Figure 5. Bar plots o relative requencies o C :N


21/27


Consider Figure 8 or the interaction C :V S ; the upper panelrepresents the interlanguage data, the lower panel represents the native speaker data,and the bars are sorted rom large absolute pairwise differences (lef) to small absolutepairwise differences (right).

Te learners and the native speakers differ most strongly with semantically moreabstract verbs and time/place verbs, as in He thinks that if he can achieve one impossi-ble act, then this will change everything .

Te learners pre er can with abstract verbs more strongly than the native speak-ers, but they pre er may more strongly with time/place verbs. However, there are also(less pronounced) differences or verbs that would typically have a human agent.

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

pl

Interlanguage

sg

can

may

can

may

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

pl

Native

sg

can

may

can

may

Figure 6. Bar plots o relative requencies o C :S N

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Animate

Interlanguage

Inanimate

can

may

can

may

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

Animate

Native

Inanimate

can

may

can

may

Figure 7. Bar plots o relative requencies o C :R A


22/27


For instance, the learners pre er may with communication verbs and can with ac-tion-trans ormation verbs. Virtually no difference at all is ound with copulas.

As or the nal interaction, C :A , we do not represent it heregraphically. While it is signicant, the large number o categories plus the act that themost pronounced differences occur with a small number o very in requent categoriesdoes not yield much in terms o interesting ndings.

As or the main effects, we will not discuss them here in detail. Tis is becausethese main effects by denition do not tell us anything about the can and may varia-bles across languages (since these variables do not interact with C ). However,since they do tell us something about which modal verb is pre erred by both nativespeakers and learners, we summarize them here visually in Figure 9. Te x -axis lists

the main effects, on the y -axis we show the percentage o can obtained or levels othese main effects, and then the levels are plotted at their observed percentage o can;the dashed line represents the overall percentage o can in the data.

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

abstr

Interlanguage

templ comm act_transf act_gen/mot ment/perc copula

can

may

can

may

can

may

can can

may

can

may

can

may

can

may

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

abstr

Native

templ comm act_transf act_gen/mot ment/perc copula

can

may

can

may

can

may

can

may

can

may

can

may

can

may

can

may

Figure 8. Bar plots o relative requencies o C :V S


23/27


Finally, a brie look at the regression’s misclassications seems to indicate thatthey did not occur randomly. While all 34 misclassications occurred in the inter-language data, 29 o them occurred with may in a orm characteristic only o theFrench-English learner language. In the large majority o those misclassications,

may is ound to express a possibility that results rom some sort o theoretical demon-stration. Consider the examples in (7) and (8). While the ones in (7) illustrate ourcurrent point, (8) provides an additional example o an atypical occurrence o learnermay , which clearly denotes a strong sense o possibility and whose interpretation isheavily reminiscent o that o can.

(7) a. So we may say that … b. o conclude , we may say that … c. As a conclusion , we may say that …

d. Tis is why we may now speak o the stupe ying effecte. Tis is the reason why we may say that …

(8) “Dresden is an old town”, we may read o its history

. Concluding remarks

By way o a summary, the BP approach and the subsequent logistic regression allows

us to recognize how can and may (in native and learner English), as well as pouvoir ,relate to each other as well as what helps determine native speakers’ and learners’choices. On the whole, distributionally we do nd the expected groupings: the cans,then the may s, and only then pouvoir . However, it is interesting that, semantically,

1 . 0

0 . 8

0 . 6

0 . 4

0 . 2

0 . 0

corpus speakerpresence

grm.accept

voice use sent.type

elliptic verbtype

subj.pers.

subj.morph.

P e r c e n

t a g e o

f c a n

strong

weak no

medium

yesilnative

passive

active

metaphorical

literal

interrogative yes

declarative no accmp-achvproos

twoone

three

state rel-dem/prother

com_noun

prop_nounpr

Figure 9. Main effects o the logistic regression


24/27


English can is more similar to French pouvoir than to English may , and the subse-quent regression results provided some initial in ormation on why that is so. More

specically, the way learners choose one o the two verbs is ofen compatible witha processing-based account in terms o the complexity principle – they choose themore basic and requent can over may when the environment is complex – but is alsostrongly inuenced by the animacy o the subject and the semantics o the verb: can isoverpre erred by learners with animate subjects and with abstract verbs, and under-pre erred with time/place verb semantics.

With regard to the modals per se, our results conrm previous studies’ recogni-tion o the inuential role o the linguistic context in the uses o may and can. Indeed,while the main effects included in our nal logistic regression model support studiesthat have identied morphosyntactic components such as V and S asparticularly inuential categories (Leech 1969, 2004; Huddleston 2002; Collins 2009),our results reveal the necessity to also take the semantic context o modals more seri-ously, as reected by the strong effects o V and V S .

More generally speaking and in the parlance o the Competition Model, the clus-ter analysis and the high classication accuracy o the regression suggest that, on thewhole, the learners have built up mental categories or can and may that are internallyrather coherent. However, the interactions in the regression show that these cues areweighted incorrectly and sometimes trigger a verb choice that is not in line with na-tive speaker choices, but that even this kind o incorrect choice is largely predictable(because the regression can still make the correct classications (c . Deshors 2010 ormore detailed discussion as well as a distinctive collexeme analysis revealing addi-tional verb-specic pre erences). In other words, even though this is the rst studyinvolving learner data (and only the second involving different languages), the BP ap-proach and especially the ollow-up in terms o the logistic regression are there ore aninteresting diagnostic: (i) the overall results can testi y to the strength o the categoriesthat are being studied, and (ii) the regression with its inclusion o the interactions o

all variables with “native speaker vs. learner” exactly pinpoints where interactionsbecome signicant, i.e. where the categories o the learner are still substantially di -

erent rom the native speaker. For urther applications and extensions, see Gries andWulff (2013) or a similar application to the choice o (of - and s-) genitives by nativespeakers and learners, and Gries and Deshors (to appear) or an even more advancedapproach to precisely pinpoint where non-native speakers’ choices deviate rom thoseo native speakers and how much so. Needless to say, more and more rigorous test-ing is necessary, but to our knowledge this is the rst study proposing this kind oapproach more generally and the use o a regression with a native-learner variable

as a measure o L2 “prociency”; the results illustrate that learners’ “non-nativeness”mani ests itsel at all linguistic levels simultaneously.


25/27


References

Aijmer, K. (2002). Modality in advanced Swedish learners’ written interlanguage. In S. Granger,J. Hung, & S. Petch- yson (Eds.), Computer learner corpora, second language acquisitionand foreign language teaching (pp. 55–76). Amsterdam: John Benjamins.

Arppe, A. (2008). Univariate, bivariate and multivariate methods in corpus-based lexicogra-phy: A study o synonymy. Unpublished PhD dissertation, University o Helsinki. Availa-ble at: .

Bartning, I. (2009). Te advanced learner variety: 10 years later. In E. Labeau, & F. Myles (Eds.),Te advanced learner variety: Te case of French (pp. 11–40). Frank urt/Main: Peter Lang.

Bates, E., & MacWhinney, B. (1982). Functionalist approaches to grammar. In E. Wanner, &L. R. Gleitman (Eds.), Language acquisition: Te state of the art (pp. 173–218). Cambridge:

Cambridge University Press.Bates, E., & MacWhinney, B. (1989). Functionalism and the competition model. In B. Mac-

Whinney, & E. Bates (Eds.), Te cross-linguistic study of sentence processing (pp. 3–73).Cambridge: Cambridge University Press.

Bybee, J., & Fleischman, S. (1995). Modality in language and discourse. Amsterdam: JohnBenjamins. DOI: 10.1075/tsl.32

Byloo, P. (2009). Modality and negation: A corpus-based study. Unpublished PhD dissertation,University o Antwerp.

Coates, J. (1980). On the non-equivalence o may and can. Lingua, 50(3), 209–220.DOI: 10.1016/0024-3841(80)90026-1

Coates, J. (1983). Te semantics of the modal auxiliaries . London: Croom Helm.Collins, P. (2009). Modals and quasi modals in English. Amsterdam: Rodopi.De Haan, F. (1997). Te interaction of modality and negation: A typological study . New York:

Garland.Depraetere, I., & Reed, S. (2006). Mood and modality in English. In B. Aarts, & A. MacMahon

(Eds.), Te handbook of English linguistics (pp. 268–287). London: Blackwell.Deshors, S. C. (2010). A multi actorial study o the uses o may and can in French-English

interlanguage. Unpublished PhD dissertation, University o Sussex.Divjak, D. S., & Gries, St. T. (2006). Ways o trying in Russian: Clustering behavioral proles.

Corpus Linguistics and Linguistic Teory , 2(1), 23–60. DOI: 10.1515/CLL .2006.002Divjak, D. S., & Gries, St. T. (2008). Clusters in the mind? Converging evidence rom near

synonymy in Russian. Te Mental Lexicon , 3(2), 188–213. DOI: 10.1075/ml.3.2.03div Divjak, D. S., & Gries, St. T. (2009). Corpus-based cognitive semantics: A contrastive study

o phasal verbs in English and Russian. In K. Dziwirek, & B. Lewandowska- omaszczyk(Eds.), Studies in cognitive corpus linguistics (pp. 273–296). Frank urt/Main: Peter Lang.

Gabrielatos, C., & Sarmento, S. (2006). Central modals in an aviation corpus: Frequency anddistribution. Letras de Hoje, 41(2), 215–240.

Gass, S. (1996). Second language acquisition and linguistic theory: Te role o languagetrans er. In W. C. Ritchie, & . K. Bhatia (Eds.), Handbook of second language acquisition (pp. 317–340). San Diego: Academic Press.

Gries, St. T. (2006). Corpus-based methods and cognitive semantics: Te many meanings oto run . In St. T. Gries, & A. Ste anowitsch (Eds.), Corpora in cognitive linguistics: Cor- pus-based approaches to syntax and lexis (pp. 57–99). Berlin: Mouton de Gruyter.DOI: 10.1515/9783110197709

http://urn.fi/URN:ISBN:978-952-10-5175-3http://dx.doi.org/10.1075/ml.3.2.03divhttp://dx.doi.org/10.1515/9783110197709http://dx.doi.org/10.1075/ml.5.3.04grihttp://dx.doi.org/10.1515/9783110226423.331http://dx.doi.org/10.1111/0023-8333.00118http://dx.doi.org/10.1111/0023-8333.00118http://dx.doi.org/10.1515/9783110226423.331http://dx.doi.org/10.1075/ml.5.3.04grihttp://dx.doi.org/10.1515/9783110197709http://dx.doi.org/10.1075/ml.3.2.03divhttp://urn.fi/URN:ISBN:978-952-10-5175-3


26/27


Gries, St. T. (2010a). Behavioural Proles 1.01: A program or R 2.7.1 and higher.Gries, St. T. (2010b). Behavioral proles: A ne-grained and quantitative approach in cor-

pus-based lexical semantics. Te Mental Lexicon , 5(3), 323–346.Gries, St. T., & Deshors, S. C. ( o appear). Using regressions to explore deviations betweencorpus data and a standard/target: two suggestions. Corpora.

Gries, St. T., & Divjak, D. S. (2009). Behavioral proles: A corpus-based approach to cognitivesemantic analysis. In V. Evans, & S. Pourcel (Eds.), New directions in cognitive linguistics (pp. 57–75). Amsterdam: John Benjamins.

Gries, St. T., & Divjak, D. S. (2010). Quantitative approaches in usage-based cognitive se-mantics: Myths, erroneous assumptions, and a proposal. In D. Glynn, & K. Fischer (Eds.),Quantitative cognitive semantics: Corpus-driven approaches (pp. 333–354). Berlin: Moutonde Gruyter.

Gries, St. T., & Otani, N. (2010). Behavioral proles: A corpus-based perspective on synony-my and antonymy. ICAME Journal , 34, 121–150.Gries, St. T., &Wulff, S. (2013). Te genitive alternation in Chinese and German ESL learners:

owards a multi actorial notion o context in learner corpus research. International Jour-nal of Corpus Linguistics, 18(3), 327–356.

Hermerén, L. (1978). On Modality in English: A study of the semantics of the modals. Lund:LiberLäromedel/Gleerups.

Huddleston, R. D. (2002). Te Cambridge grammar of the English language . Cambridge:Cambridge University Press.

Hyltenstam, K., Bartning I., & Fant L. (2005). High Level Prociency in Second Language Use.

Research program or Riksbanken Jubileums ond. (Stockholm university) http://www. biling.su.se/~AAA.

Kilborn, K., & Ito, . (1989). Sentence processing strategies in adult bilinguals. In B. Mac-Whinney, & E. Bates (Eds.), Te cross-linguistic study of sentence processing (pp. 257–291).Cambridge: Cambridge University Press.

Klinge, A., & Müller, H. H. (2005). Modality: Intrigue and inspiration. In A. Klinge, & H. H.Müller (Eds.), Modality studies in form and function (pp. 1–4). London: Equinox.

Leech, G. (1969). owards a semantic description of English. Bloomington, IN: Indiana Univer-sity Press.

Leech, G. (2004). Meaning and the English verb. London & New York: Longman.

MacWhinney, B. (2004). A unied model o language acquisition. Retrieved rom [Accessed 18 June 2010].

Neff, J., Da ouz, E., Herrera H., Martínez, F., & Rica, J. P. (2003). Contrasting the use o learn-er corpora: Te use o modal and reporting verbs in the expression o writer stance. InS. Granger, & S. Petch- yson (Eds.), Extending the scope of corpus-based research: Newapplications, new challenges (pp. 211–230). Amsterdam: Rodopi.

Nuyts, J. (2006). Modality: Overview and linguistic issues. In W. Frawley (Ed.), Te expressionof modality (pp. 1–26). Berlin: Mouton de Gruyter.

Palmer, F. (1979). Modality and the English modals. London & New York: Longman.Radden, G. (2007). Interaction o modality and negation. In W. Chłopicki, A. Pawelec, &

A. Pokojska (Eds.), Cognition in language: Volume in Honour of Professor Elżbieta aba-kowska (pp. 224–254). Kraków: ertium.

R Development Core eam (2010). R: A language and environment for statistical computing.Foundation for statistical computing . Vienna, Austria. < http://www. R-project.org >.

http://www.biling.su.se/~AAAhttp://www.biling.su.se/~AAAhttp://psyling.psy.cmu.edu/papers/CM-general/unified.pdfhttp://psyling.psy.cmu.edu/papers/CM-general/unified.pdfhttp://www.%20r-project.org/http://www.%20r-project.org/http://psyling.psy.cmu.edu/papers/CM-general/unified.pdfhttp://psyling.psy.cmu.edu/papers/CM-general/unified.pdfhttp://www.biling.su.se/~AAAhttp://www.biling.su.se/~AAA


27/27


Salkie, R. (2000). Corpus linguistics: A brie guide to research in French language and linguis-tics. AFLS Cahiers, 6, 44–52.

Salkie, R. (2004). owards a non-unitary analysis o modality. In L. Gournay, & J.-M. Merle(Eds.), Contrastes: mélanges offerts à Jacqueline Guillemin-Flescher (pp. 169–182). Paris:Ophrys.

Vendler, Z. (1967). Verbs and times. In Z. Vendler (Ed.), Linguistics in philosophy (pp. 97–121).New York: Cornell University Press.

Deshors & Gries, A Case for the Multifactorial Assessment of Learner Language. the Uses of May and...

Documents

Transcript of Deshors & Gries, A Case for the Multifactorial Assessment of Learner Language. the Uses of May and...