[PDF] Initiation au béton armé Détermination de ferraillage complet d 'une
[PDF] Les moments de force - Elemo
[PDF] Le poids d 'un objet - Qu 'est-ce que dépannez-vous?
[PDF] CALCULER UN PRIX PSYCHOLOGIQUE
[PDF] L 'analyse coût-utilité - (CHU) de Nantes
[PDF] L 'analyse coût-utilité - (CHU) de Nantes
[PDF] Calcul et structure de la molécule d 'ADN 1
[PDF] Sphère et boule - Académie de Nancy-Metz
[PDF] 21 Pyramides écologiques
[PDF] 21 Pyramides écologiques
[PDF] la rémunération des expatriés - la librairie RH
[PDF] Statistiques - Mathadoc
[PDF] Méthodologie : Comment associer une date ? un siècle et - Free
[PDF] Outil 1 Indicateurs RH et d 'activité - MDEF
[PDF] CODE DU TIMBRE
Perception&Psychophysics
1990.48 (2).1(1)-178
Interactionamongauditorydimensions:
Timbre,pitch,andloudness
ROBERTD.MELARAandLAWRENCEE.MARKS
JohnB. PierceFoundationLaboratoryandYaleUniversity,NewHaven,Connecticut
In twoexperiments,weexamined
whetheror notpairsofauditorydimensions-timbre-loudness sifiedvaluesfrom onedimensionwhiletheotherdimensionwas(1)heldconstant(baseline), (2)variedorthogonally(filtering),or (3)correlatedlinearly.
ThesubjectsshowedsubstantialGarner
interferencewhenclassifyingalldimensions-thatis, poorperformanceatfilteringrelativeto baseline. line) whencorrelatedpositively, butredundancyloss (i.e.,interference)whencorrelatednega tively. pairsofdimensionsshowedsubstantialeffectsofcongruity:Attributesfrom onedimensionwere classified areinterpretedintermsof aninteractivemultichannel modelofauditoryprocessing. How do perceptual systems analyze input from separate dimensions of sensory stimulation?The present study rep resents a continuation of our investigations into the mech anisms
1989b; Melara &Marks,1990a, I990b,
199Oc,in press;
Melara
&O'Brien,1987). Guiding ourresearchis a general model of how any set of multidimensional stimuli is processed, regardless ofwhether stimulitriggervisual, auditory, or other sensory receptors. Specifically, we con ceive of stimulusprocessingin terms of a bank of higher orderfilters orchannels.As we discusselsewhere(see Melara & Marks, 199Oc),each channel isassociated with a psychologically meaningful dimension that we call the primary dimension.ISuch primary perceptual dimensions vary maximallywith changes inparticular physicaldimen sions(e.g.,frequency,intensity, or duty cycle). Thus, in audition,separatechannels code changes in pitch (fre quency), loudness (intensity), and timbre (duty cycle), in addition to changes occurring along other dimensions such as duration.
Because ofperceptualprimacy,theperceiveris able,
through attentionalcontrol,to select and then access the results of analyses in a given channel(e.g.,for purposes ofclassification oridentification).We suggest that chan nel identification(e.g.,selectingthe pitchchannel to make frequencydiscriminations)is reasonablyaccuratefor all Preparation of this article was supportedby NationalInstitutesof Health (NIH) PostdoctoralFellowshipHD07l37and NIH Grant NS28617 to Robert Melara and by NIH Grant DC00271 and National Science Foun dation Grant BNS84-20017 to Lawrence Marks. We thank Edward Carterette,Bonnie Potts. and two anonymousreviewersfor their com ments on an earlier version. We also thank Ed Washecka, Trex Proffitt, and AlisonWittenbergfor their assistance in collecting and analyzing the data. Address allcorrespondenceto Robert D. Melara, who is now in theDepartmentofPsychologicalSciences at PurdueUniversity, West Lafayette, IN 47907.primarydimensions,independentofchangesoccurring in other channels. Theimportantfactor in multidimen sional processing, we hypothesize, is how the output from one channel(e.g.,pitch) is weighted by output from other channels(e.g.,loudness). A given pair of dimensions may beeitherseparableor interacting(seeGarner,1974, 1981; Lockhead, 1966,
1972, 1979).
Iftwo dimensions areseparable,subjects
are able to attend selectively toeitherdimension, ignor ing orthogonal variation on the irrelevant dimension. Cir cle sizeand diameter orientation are an example of separa ble dimensions(Garner&Felfoldy,1970). In our view, withseparabledimensions,the output received from the selected channel (dimension) is perceptually independent (cf. Ashby&Townsend,1986) of the output from the irrelevantchannel. This situation is depicted graphically for thedimensionsof size andorientationin panel A of
Figure 1.
Iftwo dimensions interact,orthogonalvariation on the irrelevantdimension causes
Gamerinterference(see
Pomerantz,1983, 1986; Pomerantz, Pristach,
&Carson,
1989) when the relevant dimension is being classified.
Interactionisthereforeindicated by a failure of selective attention (see,e.g.,Garner, 1974, 1981). We stress, how ever,that althoughperceiversfail to attend selectively to interactingdimensions,they can nonetheless select the appropriatechannel forclassification(i.e.,channel iden tification is accurate).Saturationand brightness of col ors are an exampleofinteractingdimensions(Gamer&
Felfoldy,1970).
We hypothesize that, with interactingdimensions, chan nels are linked at some level (or levels) of information processing,be that levelsensory/perceptual,phonemic/ graphemic,lexical, or semantic.Interactionistherefore defined by us as the leakage orcrosstalkbetween chan nels at aparticularlevel of processing of each dimension. 169
Copyright1990PsychonomicSociety, Inc.
170MELARAANDMARKS
Melara
&Marks,1990a).Crosstalkmayoccureither withina single level ofprocessingorbetweentwodiffer ent levels. Anexample ofbidirectionalcrosstalkwithin thesensory/perceptuallevel is depicted forsaturationand brightnessin panel B ofFigure1.
Aredimensions
ofsoundseparableorinteracting?Re searchon thisquestionhas been confinedprimarilyto the auditory dimensions of pitch and loudness.Evidenceindi cates that pitch and loudness do interact:
Inspeeded classi
Gamer interferencewhen pitch andloudnessarevariedortho gonally (Grau &Kemler-Nelson,1988;Melara&Marks, when these two dimensions are correlated(e.g.,high pitch pairedwith loud and low pitchpairedwith soft); that is,classificationspeed andaccuracyshow a redundancy gain (Grau&Kemler-Nelson,1988;Melara&Marks,
199Oc). Theseresults
areconsistentwith the view that (see Grau &Kemler-Nelson,1988;Melara&Marks,
199Oc;Wood,1975; seeFigure2).
Relatively little research
hasbeen devoted toward under of auditorytimbre(but seeIverson&Krumhansl,1989).
Timbreis the tonalquality
ofcomplexsounds.Itis the characteristicof sound thatpermitsus todistinguisha guitarnote from apianonoteplayedat the same fun damentalfrequencyandintensity.The lack ofresearch aimed at evaluating timbre classification reflects,perhaps, theinherentdifficulty inadequatelydefining values along thisdimension(seebelow).Nonetheless,a gappresently exists in ourunderstanding ofmultidimensionalprocess ing with thethreeprimarydimensions ofsound-timbre, pitch,and loudness. Thepresentstudyrepresentsourini tialattemptto close this gap byexaminingtimbreclas sificationwhen thatdimensionispairedwitheitherpitch orloudness.
Howperceiversprocess
timbre-specifically,whether timbreinteractswith or isseparablefromdimensions of pitchandloudness-bearsheavilyon thedevelopmentof exist. One is thattimbreinteractsbidirectionallywith both dimensions.Here, onemightenvision a triplex ofparallel channels,interconnectedat one level (orperhapsmore); later,we shalldiscuss justsuch a model.Anotherpossi bilityis thattimbreisseparablefrom one or bothdimen sions.Evidenceforseparabilitycould suggest that the au ditorysystem carries outanalysessequentially onseparate dimensions,as mightoccur ifprofileanalyses (cf.Green,
1988) followedencoding
offrequencyandamplitude.A thirdpossibilityis thatcrosstalkisunidirectional;such aresultcould also beinterpretedtoreflectdimensional hierarchiesin auditoryprocessing.Discriminationamong these ships amongtimbre,pitch,and loudness-hasobvious
Decisional
---T-
Semantic
Decisional
level -r-
Semantic
Sensory/perceptual
j;l orientation
Decisional
---T-
Semantic
Decisional
---T-
Semantic
B. A.
Sensory/perceptual
size
Sensory/perceptual-.Sensory/perceptual
j;lj;l saturationbrihtness Figure 1. Pictorialinterpretationof processesunderlyingthe sepa rability of circle sizeanddiameterorientation(A)and those under lying theintegralityofsaturationandbrightness (B).Arrows indi cate the hypothesized flowof processing from one levelof analysis toanother,eitherduring analyses of anyonedimension or when crosstalk occurs between two dimensions. Labels have been masked for the linguistic analyses (l.e., lexical access level, phonemicl ing theseperceptualdimensions.
Accordingtoourmodel, thepartialoutputfrom the ir
values(attributes)arederivedas finaloutputofprocess ing in theselectedchannel(Melara &Marks,1990b).
Gamerinterferencemay beeithersymmetric(present
wheneitherdimensionis beingclassified)orasymmetric (presentfor only a singledimension),dependingon what may beeitherbidirectional (mutual) orunidirectional(see implicationsfor thedevelopmentofpsychologicaland psychophysical models ofbroadbandsignal analysis(e.g.,
Durlach,Braida,&Ito, 1986).
A finalreasonforexaminingauditoryinteractionsis
topursuefurther ourdistinctionbetweensingle context anddual context(Melara&Marks,1990b),and the differentialeffectsthesetwocontextshave onclassifica tion. Singlecontextoccurswhen theirrelevantdimension isvariedunderasingleset ofconstraints.Forexample, varyingtwodimensionsorthogonally,as in "filtering" classcontext,"becausestimuli vary within each response category.Dualcontextoccurswhen twoforms ofvaria tion arecombined.Thus,if stimuli varywithineachcat egory(intraclasscontext)but, inaddition,therelevant and irrelevant dimensions arecorrelated("redundantcon text"),a dualcontextiscreated.
We have foundinterestingdifferencesbetweensingle
contextsand dualcontextsin theclassification ofaudi torydimensions.Whenpitch andloudnessarepaired,for example,judgmentson both dimensions showinterference withsingle(intraclass)contextualvariation.With dual only loudness classifications are worse than baseline; pitch judgmentsdo notsufferfromintraclasscontext,but show benefitfromredundantcontext.On the basis ofthis evi dence,we cal1 pitch a
HARDdimension,becausein dual
contextintraclassinterferencecan beavoided,and we call loudnessa
SOFfdimension,because in dualcontextintra
classinterferenceisinevitable.
Results from dual context also identifytimbreas a
HARD dimension:Classificationofthis dimension shows no inter ferencewhen thedimensionispairedwitheitherpitch orloudness(Melara &Marks,1990b). But note that the absence ofinterferencecan also obtain iftimbreis separa ble from pitch andloudness.Thus,it iscriticaltoevaluquotesdbs_dbs18.pdfusesText_24