Beyond the Lemma: Inflection-Specific Constructions in English Sally Rice and John Newman University of Alberta AACL 2008 BYU 14 March 2008

Bertrand Russell’s Emotive Conjugations I’m tenacious you’re stubborn he’s pigheaded singular

inflection-specific meaning and behavior (idiosyncracies of meaning, form, collocation, genre, and distribution)

Form Idiosyncracies in a Typical Athapaskan Verb Paradigm

Form Idiosyncracies in another Typical Athapaskan Verb Paradigm

TAM-Based Idiosyncracies of some Basic Verbs Rice & Newman 2005

INFPRESPASTPROGPERF 1.SG I need to go I go I went I am/was going I have/had gone 2 you need to go you go you went you are/were going you have/ had gone 3.SG s/he/it need to go s/he/it goes s/he/it went s/he/it is/was going s/he/it has/had gone 1.PL we need to go we go we went we are/ were going we have/had gone 3.PL they need to go they go they went they are/were going they have/had gone GO SUBJ x TAM Idiosyncracies in BNC all Rice & Newman 2005

GO INFPRESPASTPROGPERF 1.SG 6 % 3 % 2 % 4 % 0 % 2 2 % 10 % 0 % 3 % 0 % 3.SG 13 % 4 % 10 % 15 % 2 % 1.PL 6 % 0 % 1 % 2 % 0 % 3.PL 2 % 6 % 1 % 2 % 3 % SUBJ x TAM Idiosyncracies in BNC all Rice & Newman 2005

GO INFPRESPASTPROGPERF 1.SG 6 % 3 % 2 % 4 % 0 % 2 2 % 10 % 0 % 3 % 0 % 3.SG 13 % 4 % 10 % 15 % 2 % 1.PL 6 % 0 % 1 % 2 % 0 % 3.PL 2 % 6 % 1 % 2 % 3 % SUBJ x TAM Idiosyncracies in BNC all Rice & Newman 2005

INFPRESPASTPROGPERF 1.SG I need to think I think I thought I am/was thinking I have/had thought 2 you need to think you think you thought you are/were thinking you have/ had thought 3.SG s/he/it needs to think s/he/it thinks s/he/it thought s/he/it is/was thinking s/he/it has/had thought 1.PL we need to think we think we thought we are/ were thinking we have/had thought 3.PL they need to think they think they thought they are/were thinking they have/had thought THINK SUBJ x TAM Idiosyncracies in BNC CC Rice & Newman 2005

INFPRESPASTPROGPERF 1.SG 80 % 93 % 82 % 65 % 75 % 2 10 % 2 % 8 % 11 % 3.SG 7 % 0 % 5 % 9 % 10 % 1.PL 2 % 7 % 12 % 0 % 3.PL 1 % 3 % 4 % 6 % 4 % THINK SUBJ x TAM Idiosyncracies in BNC CC Rice & Newman 2005

INFPRESPASTPROGPERF 1.SG 80 % 93 % 82 % 65 % 75 % 2 10 % 2 % 8 % 11 % 3.SG 7 % 0 % 5 % 9 % 10 % 1.PL 2 % 7 % 12 % 0 % 3.PL 1 % 3 % 4 % 6 % 4 % THINK SUBJ x TAM Idiosyncracies in BNC CC Rice & Newman 2005

rid allow VVB-base VVZ-3sg.pres VVI-inf VVD-past VVG-prog VVN-perf part

rid allow VVB-base VVZ-3sg.pres VVI-inf VVD-past VVG-prog VVN-perf part

Inflectional idiosyncracies of EAT and DRINK Newman & Rice 2003

Distributional idiosyncracies of A, Aer, Aest Newman & Rice 2006

Collocational idiosyncracies of A, Aer, Aest Newman & Rice 2006

Really idiosyncratic gradable As rath(e) rather rathest nighnear next nearest formeformer first foremost latelater last latest

rath(e) rather rathest nighnear next nearest formeformer first foremost latelater last latest Really idiosyncratic gradable As

inflection-specific meaning and behavior (idiosyncracies of meaning, form, collocation, genre, and distribution)  Commitment to usage-based approaches like CG, RCG language-specific construction-specific inflection-specific Application of corpus linguistic methodologies and mindset

Inflectional Islands Syntactic (constructional), semantic, and collocational properties tend to inhere in individual inflections of a lexical item in a register-specific manner. These properties may not extend across all the inflections (the paradigm) to characterize the lemma as a whole.

English Pronominal Inflection searched BNC and CAE with Mark Davies’ corpus tool: Variation in English Words and Phrases: tracked person & case distribution/skewfor pronouns examined frequencies and collocations

English Pronominal Inflection NOMACC/POSSINDREFLEX OBLDETPOSS 1SG Imemyminemyself 2 youyouyouryoursyourself/ves 3SG.M hehimhishishimself 3SG.F sheherherhersherself 3SG.N itititsitsitself 1PL weusouroursourself/ves 3PL theythemtheirtheirsthemself/ves

English Pronominal Inflection NOMACC/POSSINDREFLEX OBLDETPOSS 1SG Imemyminemyself 2 youyouyouryoursyourself/ves 3SG.M hehimhishishimself 3SG.F sheherherhersherself 3SG.N itititsitsitself 1PL weusouroursourself/ves 3PL theythemtheirtheirsthemself/ves

English Pronominal Inflection NOMACC/POSSINDREFLEX OBLDETPOSS 1SG Imemyminemyself 2 youyouyouryoursyourself/ves 3SG.M hehimhishishimself 3SG.F sheherherhersherself 3SG.N itititsitsitself 1PL weusouroursourself/ves 3PL theythemtheirtheirsthemself/ves

Top 20 Collocates for Pro + V (=NOM)

Percent Attraction of Pro to Pro + CAN Frame Attraction of “he” = (no. of “he+CAN” in construction over total no. of “he” in corpus) x100, following Schmid 2000)

Percent Attraction of Pro to Pro + WILL Frame Attraction of “he” = (no. of “he+WILL” in construction over total no. of “he” in corpus) x100, following Schmid 2000)

Collostructional Analysis of 3SG + CAN/WILL Collostructional analysis based on attraction/repulsion of “she/he/it” to the construction “Pro + CAN” or “Pro + WILL” [size of corpus = number of verbs (V*)]; Stefanowitsch & Gries 2003

PRO like to vs. PRO hope to (BNC-cc) you like to I hope to

PRO like to vs. PRO hope to (BNC-cc) you like to I hope to

English Pronominal Inflection NOMACC/POSSINDREFLEX OBLDETPOSS 1SG Imemyminemyself 2 youyouyouryoursyourself/ves 3SG.M hehimhishishimself 3SG.F sheherherhersherself 3SG.N itititsitsitself 1PL weusouroursourself/ves 3PL theythemtheirtheirsthemself/ves

Top 20 Collocates for V + Pro (=ACC?)

Top 10 Collocates for P + Pro (=OBL)

Distributional Idiosyncracies (CAE all + BNC all ) prep + PRO Reliance of “to+me” (= no. of “to+me” in corpus over total no. of PREP+PRO in corpus) x100, following Schmid 2000)

English Pronominal Inflection NOMACC/POSSINDREFLEX OBLDETPOSS 1SG Imemyminemyself 2 youyouyouryoursyourself/ves 3SG.M hehimhishishimself 3SG.F sheherherhersherself 3SG.N itititsitsitself 1PL weusouroursourself/ves 3PL theythemtheirtheirsthemself/ves

Top 20 Collocates for Pro’s NP (=GEN)

Distributional Idiosyncracies (AEC all + BNC all ) my KINTERM Reliance of “my+mother” (= no. of “my+mother” in corpus over total no. of PRO.POSS+NN* in corpus) x100, following Schmid 2000)

3SG Distributional Idiosyncracies (BNC all )

1SG Distributional Idiosyncracies by Genre

1PL Distributional Idiosyncracies by Genre

Our Manifesto de-lemmatize! inflected forms have a life of their own (Tao 2001, 2003) put lemmas aside (as done earlier with syntactic rule in favor of constructions) substitute words-in-context or WICs (intersection of genre, register, & inflection) aim low! find the “hierarchy of lower-level structures...[that] specify the actual array of subcases and specific instances that support and give rise to the higher-level generalization” RWL, Concept, Image, & Symbol, 1991:

Thank you. ualberta.ca ualberta.ca

children tend to use uninflected verb roots before inflected forms verb inflections are mastered on a verb-by-verb basis generalization is gradual initially, particular verbs “strand” inflections adults use particular inflected forms of individual verbs on a register-specific basis verb inflections adhere to verbs on a verb-by-verb basis particularization is gradual eventually, inflections “strand” particular verbs THE VERB ISLAND HYPOTHESIS Tomasello 1992, 2004 THE INFLECTIONAL ISLAND HYPOTHESIS Rice & Newman 2005 V < < < inflection V > > > inflection

The Inflectional Island Hypothesis Rice & Newman 2005 uneven distribution of inflection lexical items may have “weighted” inflectional profiles weightings may be universal (experientially motivated) or language- specific inflectional categories are lexically & pragmatically meaningful (and not just part of grammatical house-keeping or concord relationships) especially “weighty” inflected items (WICs) may idiomaticize and grammaticalize

Distributional Idiosyncracies (AEC all + BNC all ) my BODY PART

words in context (WICs) + distribution patterns (usage) collocations & N-grams pragmatic associations incipient grammaticalization & idiomaticization lemmas argument structure(s) syntactic constructions lexical meaning inflected forms “have a life of their own” Thompson & Hopper 2001:44

WICs locus of lexicalization and grammaticalization active in borrowings and morphological realignment spawn psychological associations, induce priming effects

Corpus-based constructionists approaches allow us to leave the paradigm behind. Paradigms (lemmas) have value for some purposes, but they often end up straight-jacketing an analysis. The conceit of the paradigm tends to distract linguists from looking at the lexical semantics of inflected forms in their own right. Some items do escape the shackles of the paradigm and actually become lexical items in their own right (as in Russellian conjugations).