Lexical Competition Correlates with Articulatory Enhancement in ASL Kathleen Currie Hall, Oksana Tkachman, & Yurika Aonuki

Department of Linguistics, University of British Columbia

Research Question & Hypothesis Q:Howdoeslexicalcompetitionaffecttheproductionofsigns?H:Greaterneighbourhooddensitywillcorrelatewithincreasedvisibleamplitude.

Fig.3:(a)Generalvoweldispersioncf.Munson&Solomon(2004);(b,c)Specificdistancingfromaneighbourcf.Wedeletal.(2018)andBaese-Berk&Goldrick(2009)

Methodology •  ASL-Lex(Casellietal.2017):•  Databaseof~1000signsofASL•  Articulatedbyasingledeafnativesigner•  Eachsignedinisolation

•  ASL-SignBank(Hochgesangetal.2019):•  Databaseof~2000ASLsigns•  Articulatedbydeafnativesigners•  Eachsignedinisolation•  IncludedonlysignsthatalsooccurinASL-Lex

•  “Minimal”NeighbourhoodDensity(ND)•  TakenfromASL-Lex•  Thenumberofsignsthatshareatleastoneof5characteristics

withagivensign(#ofhands,majorlocation,majormovement,selectedfingers,fingerflexion)

•  MostsimilartomeasuresusedbyothersignNDstudies•  Thought to better “capture the phonological structure of the

lexicon”(Casellietal.2017:9)•  Removed:•  Compounds,atypicalhandednessorlocation,clippedvideos

•  Totalsignsanalyzed:•  691videosforASL-Lexand644forASL-SignBank

•  Opticalflowanalysis:•  FlowAnalyzersoftware(Barbosa2013)•  Visibleamplitude(VA)ofeachsigncomputedasinFig.4

•  Linearmodel:•  Visible amplitude~numberofhands+major location+minor

location+majormovement+ND•  ComparetomodelwithoutND

Results

v1 v2

v3 v4

z1 z2z4z3

x1y1

Background: Lexical competition •  Lexicalcompetition:networksofphonologicallyrelatedwords•  Spokenlanguages:neighboursarewordswith1-sounddifferences•  Signedlanguages:neighboursarewordsthatsharesomecharacteristic(s)

References Baese-Berk,M.,&Goldrick,M.(2009).Mechanismsofinteractioninspeechproduction.LanguageandCognitiveProcesses,24,527-554.Barbosa,A.V.,Yehia,H.C.,&Vatikiotis-Bateson,E.(2008).Linguisticallyvalidmovementbehaviormeasurednon-invasively.InR.Goecke,P.Lucey,&S.Lucey(Eds.),Auditoryandvisualspeechprocessing--AVSP08(pp.173-177).MoretonIsland,Australia:CausalProductions.Caselli,N.,SevcikovaSehyr,Z.,Cohen-Goldberg,A.,&Emmorey,K.(2017).ASL-Lex:AlexicaldatabaseofAmericanSignLanguage.BehaviorResearchMethods,49(2),784-801.doi:doi:10.3758/s13428-016-0742-0Chan,K.Y.,&Vitevitch,M.S.(2009).Theinfluenceofthephonologicalneighborhoodclusteringcoefficientonspokenwordrecognition.JournalofExperimentalPsychology:HumanPerceptionandPerformance,35(6),1934-1949.Hochgesang,J.A.,Crasborn,O.,&Lillo-Martin,D.(2019).ASLSignbank.Retrievedfromhttps://aslsignbank.haskins.yale.edu/Horn,B.K.P.,&Schunck,B.G.(1981).Determiningopticalflow.ArtificialIntelligence,17,185-203.Lindblom,B.(1990).Explainingphoneticvariation:AsketchoftheH&Htheory.InW.J.Hardcastle&A.Marchal(Eds.),Speechproductionandspeechmodelling(pp.403-439).Dordrecht:Kluwer.Morgan,H.E.(2017).ThePhonologyofKenyanSignLanguage(SouthwesternDialect).(PhDDoctoraldissertation),UniversityofCalifornia,SanDiego.Munson,B.,&Solomon,N.P.(2004).Theeffectofphonologicalneighborhooddensityonvowelarticulation.JournalofSpeech,Language,andHearingResearch,47(5),1048-1058.Wedel,A.,Nelson,N.R.,&Sharp,R.(2018).Thephoneticspecificityofcontrastivehyperarticulationinnaturalspeech.JournalofMemoryandLanguage,100,61-88.doi:https://doi.org/10.1016/j.jml.2018.01.001

Fig.2:ExampleofphonologicalneighbourhoodforthesignFALSEinKenyanSignLanguage(Morgan2017:108).“FALSEhasthreetrueminimalpairs,atleastsevennear-minimals,andanuncountednumberof3-differencepairs.”

Fig.1:ExampleofphonologicalneighbourhoodforthewordspeechinEnglish,withthreeminimalpairsshownandsevennear-minimalshighlighted(Chan&Vitevitch2009).

Background: Articulatory Enhancement •  Increaseddistinctivenessofasignalismostlikelytohappenwhenthereisthe

greatestchanceofmiscommunication(cf.Lindblom1990)•  Has various effects on both recognition andproduction, though effects can

varywithdiscoursecontext,phonologicalcontext,measurementtype...:

Measuring “Visible Amplitude” (VA) •  Analogoustoacousticamplitude•  Amountofenergyproducedbymotionsthatcompriseasign•  Affectedbye.g.numberofhands,type/shapeofmovement•  CalculatedfromavideoofasignbyapplyingOpticalFlowAnalysis

(OFA;Horn&Schunck1981,Barbosaetal.2008)

Fig.4:(a)Frame1;(b)Frame2;(c)Opticalflowfield;(d)Calculatingmagnitudesofindividualvectorsinthefield.Tocalculatemagnitudeofframe-step,averagethemagnitudes(z-values)from(d).TocalculateVA,squarethemagnitudesacrossframe-steps,sumthem,dividebythenumberofframe-steps,andtakethesquareroot.

Discussion

ASL-Lex(left):•  NDisasignificantpredictor

(p=0.009)•  Effectisinexpected

direction,thoughsmall

ASL-SignBank(right):•  NDisasignificantpredictor

(p=0.015)•  Effectisinexpected

direction,thoughsmall

•  Lexicalcompetitionmayaffectarticulationinsignedlanguagesinamannersimilartothatinspokenlanguages.

•  Increasedcompetitionisassociatedwithincreasedmagnitudeofmovementsinsigns.

•  Wedeletal.(2016):spokenlanguageeffectsbettercapturedbylexical-item-specificmeasuresthangeneralizedND–howcanwecapturethisinsignedlanguages?

•  Note:asimilaranalysisofthe“maximal”NDmeasureinASL-Lex(neighboursshare4of5characteristics)showednosignificanteffectofNDineitherdatabase.