Page305
units,researchershavetraditionallyreliedonconcurrentTHINKALOUD
PROTOCOLS(TAPs),thatis,translators’ownattemptstoverbalizetheir
thoughtprocessesastheytranslate.Onedifficultywiththismethod,however,is
thatonlyasmallproportionofanyparticularTAPislikelytogiveinformation
aboutattentionalfocusonthesourcetext(Krings2001:314).Moreover,
becauseverbalizationsrelatedtothesourcetexttendtooccuronlywhen
subjectsareexperiencingparticularproblemswiththattext,someresearchers
cometoequate‘translationunit’with‘translationproblem’.LivbjergandMees
(2003:129),forexample,define‘translationunits’as‘Anywordorphrasein
the[source]text,oranyaspectofsuchawordorphrase,whichisverbalised
byanysingleparticipantandforwhichheorsheexpressesanydegreeofdoubt
aboutitspropertranslation’.BarbosaandNeiva(2003:138–9),ontheother
hand,suggestthattranslationunitsarenotsomuchdefinedbyproblemsas
demarcatedbyproblemswhichcausebreaksinthetranslation‘flow’.
Morerecently,pausedataobtainedbymeansofkeystrokelogginghasbeen
usedtoexplorethewaytranslatorssegmentorchunktheirprocessing(see,for
example,Alves2005;Jakobsen2002,2005),theassumptionbeingthat
distributionsofpausesinthetranslationprocessingstreamwouldgivean
indicationof‘howmuchisbeingprocessedatanyonetime’(Jakobsen
2005:173).
Eventhoughprocessorientedtranslationscholarsstressthattranslationunits
aredynamicandcannotbeequatedwithstructuralunitsofanyparticularrank,
somestudieshavecometoconclusionsabouttheranksatwhichtranslators
withdifferentlevelsofexpertisetendtooperate.Thus,Gerloff(1988)and
Kiraly(1990)–bothcitedinKrings(2001)–andLörscher(1991,1993)
concludethatprofessionaltranslatorstendtofocusonsourcetextunitsofhigher
rankthandosemiprofessionalsornonprofessionals,andBarbosaandNeiva
(2003:139)foundthatmoreadvancedforeignlanguagelearnerstranslatinginto
L1processlongertranslationunits,athigherlinguisticlevels,thandoless
advancedstudents.Jakobsen(2005:183)alsofoundthatexperttranslators
workwithlongersegmentsthantranslationstudentsandwhentranslatinginthe
L2toL1directionthanwhentranslatingintotheirsecondlanguage.
Theunitoftranslationinproductorientedtranslationstudies
Whileprocessorientedapproachestotranslationunitsgiveprioritytosource
textsegments,productorientedapproachesstartwithtargettextsandviewthe
unitoftranslationas‘thetargettextunitthatcanbemappedontoasource–text
unit’(Malmkjær1998a:286).Therearefewdetaileddescriptionsofhowsuch
mappingscanbecarriedout,butToury(1980a,1995)givessomerelevant
guidance.Touryisinterestedinidentifyingcoupledpairsoftargetsolutions
(‘replacing’segments)tosourceproblems(‘replaced’segments),butconcedes
thattheboundariesofsuchcoupledpairsaredifficulttodetermine,giventheir
dynamicnatureandhighcontextdependency.Hethusadvocatestheapplication
ofa‘noleftovers’principlewherebytheanalystgoesaboutestablishing‘a
segmentofthetargettext,forwhichitwouldbepossibletoclaimthat–beyond
itsboundaries–therearenoleftoversofthesolutiontoatranslationproblem
whichisrepresentedbyoneofthesourcetext’ssegments,whethersimilaror
differentinrankandscope’(1995:78–9).ForToury,andZabalbeascoa
(2000),whooffersasimilartreatment,‘problems’and‘solutions’–combined
byZabalbeascoa(ibid.)intoasingle‘translationunit’–aremutuallydefining,
dynamic,andspecifictoindividualpairsoftexts.Theyarealsoarrivedat
subjectively,andare,toacertainextent,anartefactoftheanalysis,asimplied
byTourywhenheadvisesthat‘whateverunitsonechoosestoworkwith
shouldberelevanttotheoperationwhichwouldthenbeperformedon
them’(1995:88;emphasisinoriginal).AlthoughToury’scoupledpairsserve
primarilytoassisttranslationanalystsinreconstructingtranslationdecisions,he
doessuggest(ibid.:99)thattheymayalsohaveafunctionintheimplementation
oftranslationdecisions,withtranslatorspotentiallystoringcoupledpairsinlong
termmemory,andretrievingtheminsubsequenttranslationtasks.