TEACHING LINGUISTICS Learning to think like linguists: A think-aloud study of novice phonology ...

18 Pages·2016·0.17 MB·English

TEACHINGLINGUISTICS
Learningtothinklikelinguists:Athink-aloudstudyofnovicephonologystudents
CATHERINEANDERSON
McMasterUniversity
Akeylearningoutcomeforundergraduatelinguisticscoursesisforstudentstolearntoreason
scientificallyaboutlanguage.Thisarticlepresentsthefindingsfromathink-aloudstudyofunder-
graduatesinanintroductorylinguisticscoursewhowereintheprocessoflearninglinguisticrea-
soningaboutphonology.Idescribethestudents’developingconceptsandmakerecommendations
forinstructorstohelpstudentsdevelopfullyformedlinguisticsconceptsandtheabilitytothink
scientificallyaboutlanguage.*
Keywords: pedagogy, think-aloud, threshold concepts, novice thinking, disciplinary thinking,
phonology
Every field has ways of thinking that characterize the discipline (Anderson &
Hounsell 2007, Middendorf & Pace 2004, among many others), and the field of lin-
guisticsisnodifferent.Kuiper(2011:183)assertsthat‘[w]henyougetagroupoflin-
guists together there is remarkable agreement about the domains that are central to
linguistics’.InasurveyoflinguisticsinstructorsinNorthAmericaandEurope(Ander-
son2012),Iaskedparticipantstorateasetofpossiblelearningoutcomesforanunder-
graduatelinguisticsdegree.Allseventy-onerespondentsagreedorstronglyagreedthat
students should understand the principles of scientific reasoning as they apply to the
studyoflinguistics.Acoreelementofscientificreasoninginlinguisticsistheabilityto
thinkaboutlanguageasobservabledataandtodrawconclusionsfromone’sempirical
observations. In this article, I describe a study of beginning linguistics students’ at-
tempts to make observations and draw conclusions about the phonology of two unfa-
miliarlanguages.Thefindingsprovideinsightforinstructorsintothenatureofnovice
linguists’mental representations of key linguistic concepts such as phonemic contrast
andallophonicvariation.
Becauseofthepervasivenessofprescriptiveattitudesaboutlanguage,theabilityto
threshold concept
think empirically about language is no doubt a for novice lin-
guists. Meyer and Land’s (2003) influential theory argues that threshold concepts,
whicharefoundationaltothewaysofthinkingwithineachdiscipline,aretransforma-
tional for learners, ‘opening up a new and previously inaccessible way of thinking
aboutsomething’(2003:1).Butthistransformationisdifficultandtimeconsuming,and
itcantakeweeksormonthsfornovicestocrossthethresholdintoadisciplinarywayof
thinking. During these weeks, Meyer and Land argue, students function in a liminal
space, ‘a suspended state in which understanding approximates to a kind of mimicry’
(2003:10). The present study of students in an introductory course offers a view of
novices’concepts within this liminal space, where students approach the threshold at
differentrates.Frommyfindings,Ioffersomerecommendationstoinstructorsofintro-
ductorylinguisticscoursesforsupportingstudentsincrossingthethresholdtolinguis-
ticthinking.
The think-aloud method. think-aloud method
1. The (Ericsson & Simon
1993,vanSomerenetal.1994)offersresearchersausefultoolforobservingthinking
*IgratefullyacknowledgesupportfromtheFacultyofHumanitiesatMcMasterUniversityandespecially
theworkoftwoundergraduateresearchassistants,LauraBeaudinandLeslieHumphries,whocollectedand
transcribedthedata.
e274
PrintedwiththepermissionofCatherineAnderson.©2016.
TEACHINGLINGUISTICS e275
processes. Ordinarily, mental processes are hidden not only from an observer or re-
searcher,butusuallyalsofromthethinkerherself.Themethodinvolvessimplyasking
participantstotalkoutloudabouteverythingtheythinkwhiletheyattempttocomplete
a task of some kind. The researcher audio-records the participants’ utterances, tran-
protocols
scribesthem,andanalyzestheresultingtranscripts,whicharereferredtoas
within this research tradition. The think-aloud method is useful for studying learning
for several reasons. Blackwell and colleagues (1985) give evidence that a think-aloud
exercisegeneratessignificantlymoredatathanaskingparticipantstolisttheirthoughts
retrospectively, and Genest and Turk (1981) note that statements generated via the
think-aloud method are relatively unlikely to be subject to participants’causal infer-
encesabouttheirownthoughts,unlikearetrospectivelistingofthoughts.Theprotocols
resulting from a think-aloud reveal not just the mental processes involved in the task,
butalsotheorderinwhichthoseprocessesunfoldovertime(Whitney&Budd1996).
Researchers have used the think-aloud method to investigate general problem-
solving (Newell & Simon 1972) and inferencing processes while reading texts
(Trabasso & Magliano 1996,Whitney & Budd 1996).Although the level of self-con-
sciousnessrequiredtodescribeone’sthinkingaloudcandisruptcognitiveprocessesfor
tasksthatareoverlearnedorautomatized(Genest&Turk1981),athink-aloudisideal
for observing the mental processes of beginners, since the very process of learning a
new task involves conscious reflection on how to carry out the task (Chi et al. 1994,
Ericsson et al. 1993). For that reason, researchers of teaching and learning have em-
ployedthemethodtoinvestigatehowstudentsbegintodevelopdisciplinarythinkingin
history(Wineburg1991),politicalscience(Bernstein2010),andmechanicalengineer-
ing(Miller-Young2013).Themethodhasbeenparticularlyusefulforobservingthede-
velopmentofclinicalreasoningacrossmanyfieldsinthehealthprofessions(Banning
2008,Forsbergetal.2014,Kuipers&Kassirer1984,Wainwright&McGinnis2009).
scholarship of teaching and learning
The present article extends the existing
(SoTL)literaturetoconsiderthecognitiveprocessesandconceptualrepresentationsof
novicestudentsoflinguistics.
Participants.
1.1. Asmall number of participants is usually considered sufficient
forathink-aloudstudy,ontheassumptionthatthereareonlyafewqualitativelydiffer-
ent ways that students will understand (or misunderstand) course concepts (Marton
1986).SevenMcMasterUniversityundergraduatestudents(agesseventeentotwenty-
five)participatedinthisstudy.1Sixwerenovicesenrolledinthefirstsemesterofatwo-
semesterIntroductiontoLinguisticssequence2andhadnottakenanyotherlinguistics
classes.Theseventhstudentwasafourth-yearlinguisticsmajorwhohadcompletedall
of the core theoretical classes in the program and was working as a teaching assistant
(TA)fortheIntroductioncourse.
The participants were recruited via the department’s experiment participation soft-
ware.Ascompensationtheyearnedonehourofparticipationcredittowardalinguistics
course. Two research assistants (RAs) handled recruiting and compensation and con-
ductedthesessions,sothatparticipants’identitieswereunknowntomeinmydualroles
ofresearcherandcourseinstructor.Allparticipantsgaveinformedconsenttoparticipate.
1Wedidnotcollectgenderinformationfromourparticipants.Iusefemininepronounstorefertostudents
throughoutthisarticlebecausethelargemajorityoflinguisticsstudentsatMcMasterarewomen.
2Thefirst-semestercoursedealswithphonetics,phonology,andmorphology,whilethesecondsemester
includessyntax,semantics,andpsycholinguistics.ThelanguageofinstructionatMcMasterisEnglish.
e276 LANGUAGE,VOLUME92,NUMBER4(2016)
aterials.
1.2. M Each participant attempted two exercises, reproduced in Tables 1
and2.
HereisasetofdatatranscribedfromLebaneseArabic.
[btækli] ‘youeat’ [ ntu] ‘you’(plural)
ʔɪ
[k si] ‘chair’ [bnt] ‘girl’
ʊɾ ɪ
[ nti] ‘you’(singular) [mtl] ‘like’
ʔɪ ɪɪ
[ma i] ‘withme’ [ li] ‘tome’
ʕ ʔɪ
[fi] ‘thereis’ [b i] ‘hestayed’
ɪʔ
Considerthevowels[i]and[].Aretheytwoseparatephonemes,oraretheyallophonesofonephoneme?
ɪ
Whatistheevidenceforyourconclusions?
Table
1.LebaneseArabicexercise,fromCowan&Rakušan1987.
HereisasetofdatatranscribedfromAngas,alanguagespokeninNigeria.(Rememberthatthisdiacritic
˳
indicatesthatasoundisvoiceless.Thediacriticwindicatesthataconsonantisrounded.Thesymbol[ ]
ɓ
standsforavoicedimplosivebilabialstop.)
[mut] ‘todie’ [pampam] ‘bread’
̥
[ŋ ak] ‘snake’ [ntaŋzum] ‘wasp’
ɡ ̥
[ndarm] ‘bark’ [nfwarm] ‘headcold’
̥ ̥
[nuŋ] ‘toripen’ [m m] ‘tolick’
˳ ɓɛl ̥
[mbaŋ a] ‘drum’ [tam] ‘bench’
ɡ ̥
[dondon] ‘yesterday’ [potiŋ] ‘sky’
̥ ˳
[d ŋ] ‘todrag’
ɛ˳
Considerthenasalconsonantsinthisdataset.Howmanyseparatephonemesarethere?Howmanyofthem
areallophones?Whatistheevidenceforyourconclusions?
Table
2.Angasexercise,adaptedfromHalle&Clements1983.
FivestudentsattemptedtheAngasexercisefollowedbytheLebaneseArabicexercise.
OnenovicestudentattemptedAngasfirst,movedontoattemptLebaneseArabic,then
returnedtotheAngasexercise.Onenovicestudent,thefirsttoparticipate,attemptedan
InuktitutexerciseinsteadofAngas.Sincethecourse’sTAsthenusedthatsameInukti-
tutexerciseinclass,wereplaceditwithAngasforallsubsequentparticipants,andex-
cludedthedatafromtheInuktitutexercisefromouranalysis.
Procedure
1.3. . The think-aloud sessions took place during weeks 8 to 12 of the
thirteen-week semester.The topics of phonemic contrast and allophonic variation had
beenintroducedinweek7ofthesemester.
TwoRAsconductedallsessionsinaquietroomoncampus.Atthebeginningofeach
session,participantsreadtheletterofinformation,signedtheconsentform,andfilled
outabriefparticipantinformationform.
Eachparticipantworeahead-mountedmicrophoneconnectedtoacomputerrunning
audio-recordingsoftware.TheRAsinvitedtheparticipanttothinkaloudbysayingthe
following.
(1) Pleasecompletetheexercisewhiledescribingyourthinkingoutloud.Tryto
explain what you’re thinking while you work.We will give you prompts if
you stop talking while you are working, but we cannot help you if you get
stuck.(R)3
3Excerptsfromtheprotocolsarelabeledwithacodeindicatingthespeaker:RforaResearchassistant,N
foraNovicestudent,andEfortheExperiencedstudent.
TEACHINGLINGUISTICS e277
The RAs asked each participant to read the exercise instructions aloud while they
checked the recording levels.After reading the instructions, participants began to de-
scribetheirthinking.Theyalsowroteontheexercisepaperasiftheywerecompleting
agradedassignment.Ifparticipantsstoppedspeaking,theRAspromptedthemwitha
questionoraremindertokeepthinkingoutloud.Mostpromptsweresimplereminders
tokeeptalking,forexample,‘Whatareyouthinkingaboutnow?’.Someoftheprompts
bytheRAsincludedcoachingsuchasaTAmightprovide,asin2.
(2) Goingbacktothenasalconsonants,haveyoudeterminedhowmanyseparate
phonemesthereare?(R)
Charters(2003:73)hasarguedthat,‘[i]deally,participantsinathink-aloudstudyshould
notneedanycoachingbutshouldenunciatetheirinnerspeechspontaneously’.Itispos-
siblethattheRAs’coachingpromptsmadetheprotocolunrealisticasarepresentation
ofthestudents’abilities.Inotherwords,ifthestudentwereattemptingtocompletethis
problemunassistedonanexam,shemightnotgetanyfurtherwithoutsuchaprompt.
Butmypresentinterestisnotjustinwhetherstudentscanreachtherightanswerunas-
sisted,butalsoinwhattheirmentalprocessesarewhiletheyarelearninghowtosolve
aproblem.TheRAs’occasionalcoachingpromptsgaveusafullerdatasettoillustrate
thethinkingthatwouldoccurinareal-lifelearningsituation,wherestudentsoftenre-
ceivecoachingfrompeers,TAs,andinstructors.
Ifaparticipantwasunabletoreachasolution,theRAsconcludedthesessionbyask-
ingsomeretrospectivequestionsabouttheexercise,suchas3and4.
(3) Wasthereanythingthatyoustruggledwith?(R)
(4) Whatdoyouthinkwouldbehelpfulforyoutocompletethisproblem?(R)
Thesessiondurationsrangedfromfourteentothirty-sevenminutes(averagetwenty-
four minutes).After each session the RAs transcribed the participant’s utterances into
textfiles.
Protocols and analysis.
2. The RAs transcribed seven sessions of the Lebanese
Arabic exercise (six novice, one expert) and six sessions of theAngas exercise (five
novice,oneexpert).WeexcludedtheInuktitutexercisefromthepresentanalysisinthe
interest of greater comparability across protocols. This resulted in a total of thirteen
protocols(elevennovice,twoexpert)foranalysisandcoding.
Inmycodingofthetranscripts,Iusedaninductiveapproachthatdidnotbeginwitha
hypothesisorpreconceivedsetofcodes;rather,Iallowedthecodesystemtoemergefrom
myobservationsoftheprotocols.Nevertheless,myapproachwasnottheory-neutralbe-
causemyobservationsandinferenceswereguidedbyalargeliteratureonthresholdcon-
ceptsanddisciplinarythinking(Meyer&Land2003,2005,2006,Middendorf&Pace
2004,Wismathetal.2015),bythecognitivepsychologyliteratureonproblemsolving
(Ayres1993,Catrambone1998,Chietal.1989,Ericsson2003,Getzels1982,Larkinet
al.1980,Lesgold1988,Newell&Simon1972,Owen&Sweller1985,Pretzetal.2003,
Reed1987,Reed&Bolstad1991),andofcoursebymyownexperienceofmorethana
decadeofteachingintroductorylinguistics.
Solving a phonology problem
2.1. .The typical data set from an unfamiliar lan-
guage,accompaniedbyinstructionstoconsiderapairofphoneticsegments,isgener-
ally referred to by linguists as a problem set. Although in my classes I use the term
exercise,theseexercisesareexamplesofclassicproblems:theyhavecleargoalsanda
well-definedpathtosolution.Exercisesofthiskindare,inGetzels’s(1982)terms,the
presented
typeofproblemsthatare tostudents(asopposedtodiscoveredorcreated
bytheproblem-solvers).
e278 LANGUAGE,VOLUME92,NUMBER4(2016)
Inausefulreview,Pretzandcolleagues(2003)summarizealargeliteratureonprob-
lemsolvingbylistingthecommonstagesinvolvedinsolvingaproblem.
ii(i) Recognizeoridentifytheproblem.
i(ii) Defineandrepresenttheproblemmentally.
(iii) Developasolutionstrategy.
(iv) Organize[one’s]knowledgeabouttheproblem.
i(v) Allocatementalandphysicalresourcesforsolvingtheproblem.
(vi) Monitor…progresstowardthegoal.
(vii) Evaluatethesolutionforaccuracy. (Pretzetal.2003:3–4)
Theyalsonotethatthesestagesarenotnecessarilysequential,norareallstagespresent
ineveryproblem-solvingsituation.Sincethephonologyproblemsunderconsideration
are presented to students, stage (i) of the cycle is not relevant in the present situation.
Likewise,becausethenovicestudentsaretotalbeginnersatdoinglinguistics,mostare
not yet independently capable of developing a solution strategy (stage (iii)). Instead,
most attempt to follow an algorithm or ‘recipe’learned in class. In this introductory
class,phonologyexercisesweremodeledwiththestrategyshownin5.
(5) a. Lookfirstforminimalpairs.
b. If you find minimal pairs, conclude that the segments are separate
phonemes.
c. Ifyoudon’tfindminimalpairs,considerthedistributionofthesegments.
d. If you find complementary distribution, conclude that the segments are
allophonesofonephoneme.
Asweshallseefromtheprotocoldata,many,butnotall,studentswereabletofollow
this recipe,butdidnotalwaysreachacorrectsolution.Someerrors resultedfromap-
plying the strategy incorrectly (stage (iii)), while many arose earlier, from students’
mentalrepresentationoftheproblematstage(ii).Insomecases,studentsexecutedthe
stepsaccordingtotherecipebutstillfailedtoreachacorrectsolutionbecauseofgaps
orerrorsintheorganizationoftheirknowledgeabouttheproblem(stage(iv)).
The‘correct’solutions
2.2. .TheLebaneseArabicdatasetwasthesimplerofthe
two,becauseitquestionedonlyasinglepairofsegments.Theevidenceconfirmsthat
the students found this exercise easier: four of the six novices reached a correct solu-
tion,asdidtheexperiencedstudent,andtheaveragetimetocompletethisexercisefor
thenoviceswas7;35minutes(ascomparedto12;50minutesfortheAngasexercise).
For an expert linguist, a glance at the data is sufficient to observe that [i] appears
word-finallyand[]appearselsewhere,andtoconcludethereforethat[i]and[]areal-
ɪ ɪ
lophones of one phoneme. Indeed, the experienced undergraduate participant reached
this conclusion almost immediately. Item 6 shows her first utterance after reading the
instructions.
(6) I’mobservingthedatatocheckwhatpositionsthe[i]andthe[]arefalling
ɪ
into. Initial observations are that [i] are happening only word final and that
[]ishappeningmedial,medial,medial,medial,medial.Well,they’renever
ɪ
appearing in the same environments. So, based on like a really basic basic
observation,youcouldalreadystatethatthey’reincomplementarydistribu-
tion.(E53)
Afterthreemoresentencesduringwhichsheconductedamoredetailedexaminationof
thephoneticenvironmentofeachsegment,sheconcludedasfollows.
(7) It kinda looks like they are indeed allophones of one phoneme because
they’reneverappearinginthesameenvironments.(E53)
TEACHINGLINGUISTICS e279
Foranovicelinguist,however,reachingthatconclusionwouldrequireseveralsteps.
Toreachasuccessfulsolution,thatis,onethatwouldearnfullpointsonanexam,astu-
dentwouldneedtodothefollowing.
(8) a. Activatethesolutionstrategydescribedabovein5frommemory;
b. locatetherelevantsegmentsinthedataset;
c. comparewordformstodiscoverwhetherthesegmentsappearinminimal
pairs(andinthisexercise,avoidbeingdistractedbyanirrelevantminimal
pair);
d. accessphoneticknowledgefrommemorytodescribetheenvironmentsin
whicheachsegmentoccurs(optionallyusinganenvironmentchart);
e. recognizethedistributionofthesegmentsascomplementary;
f. usethatrecognitiontoconcludethatthetwosegmentsareallophonesof
onephoneme.
TheAngasexercisedoesnotincludemuchmoredatathantheLebaneseArabicdata
set(thirteenwordsinsteadoften),butitisconceptuallymorecomplexbecauseitasks
studentstoconsidernotjustasinglepairofsegments,butthenaturalclassofnasalcon-
sonants.Thestepstosolvetheproblemareessentiallythesameasthestepsdescribed
above in 8 for LebaneseArabic, but arriving at the correct solution requires the addi-
tionalconceptualleapofrecognizingthattheentirenaturalclassparticipatesintheal-
lophonicvariation,suchthatthevoicelessnasals[mnŋ]appearword-finallyandtheir
̥ ̥ ˳
voicedcounterparts[mnŋ]appearelsewhere.
Withinthenaturalclass,therearenominimalpairsinthedatasettoprovideconclu-
sive evidence that /m/, /n/, and /ŋ/ are three separate phonemes, but there is one near-
minimalpair([mut]vs.[nuŋ])forthe/m/-/n/contrast.However,thissmalldatasetalso
˳
includessomeevidencethatcouldbeinterpretedasshowingasingleunderlyingnasal
changingitsplaceofarticulationaccordingtothefollowingconsonant,sinceinmany
ofthewordsprovided,[m]precedesbilabialsegments,[n]precedesalveolars,andthe
single instance of word-initial [ŋ] precedes [ ]. In short, the evidence for phonemic
ɡ
contrasts between [m], [n], and [ŋ] is subtle and would likely therefore be quite chal-
lengingforanovice.
Students’approachestotheproblems
2.3. .Mostofthestepssketchedabovein
8arethemselvesconceptuallyquitecomplex:theyrelynotjustonrecognitionandrecall
phonetic fea-
but on rich mental representations of core linguistic concepts such as
tures naturalclasses minimalpairs allophonicvariation
, , ,and .Chiandhercol-
laborators (1989) showed that students’abilities to solve problems arise from robust,
completeknowledgerepresentations.Indeed,thechiefgoalofmostintroductorycourses
inanydisciplineisforstudentstodevelopknowledgerepresentationscharacteristicof
thatdiscipline(Anderson&Hounsell2007,Middendorf&Pace2004,Pace&Midden-
dorf2004).Sinceourstudytargetedstudentsatthebeginningoftheirlinguisticsstudies,
weexpectedtofindthatmanynovicestudentshadfragmentaryrepresentationsoflin-
guisticconcepts.Theprotocolsfromtwonovicestudents(N84andN53)whowereun-
abletoreachacorrectsolutionforeitherexerciseindicatethatthesestudents’knowledge
representationswereconfusedandlackingdetail.Perhapsmoreinterestingtoinstructors,
theprotocolsfromsuccessfulsolutions(evenfromtheexperiencedstudent)revealsome
misconceptionsthatpersistedinstudents’conceptsevenwhentheywereabletoreach
technicallycorrectsolutions.
Representingtheproblem
.Forbothexercises,theprotocolsindicatethatstudents
spentlittletimeinformingamentalrepresentationoftheproblem(stage(ii)fromPretz
e280 LANGUAGE,VOLUME92,NUMBER4(2016)
etal.2003),andinsteadveryquicklybeganemployingthealgorithmlearnedinclass.
Eachofthefollowingquotesisastudent’sfirstutteranceafterreadingtheinstructions.
(9) So looking at the nasal consonants, how many separate phonemes of the
nasalconsonantsarethere?I’mgoingtolookattheenvironmentsthatthey
arein,soI’mgoingtomake—SoI’mgoingtoputthenasalconsonantsina
chartandthenI’mgoingtoputtheenvironmentofeach.(N51)
(10) WellI’llprobablydrawachart,likehowIstarted,butwhatdoyoumeanby
thenasalconsonants?(N39)
Evidence from a variety of domains shows that expert problem-solvers devote pro-
portionallygreateramountsoftimethannovicestosettingupamentalrepresentation
of a problem, and proportionally less time executing the solution (Chase & Simon
1973,Larkinetal.1980,Lesgoldetal.1988).Theprotocolstatementsin9and10are
examplesoftypicalnovicebehavior:thestudentsattemptedastrategywithoutcreating
aclearrepresentationoftheproblem.Infact,item10indicatesthatstudentN39hadnot
marshalledenoughphoneticsknowledgetoknowwhichsegmentstoconsiderbuthad
alreadybeguntocreateanenvironmentchart.
By contrast, students with a clear representation of the problem were able to apply
theproblem-solvingstrategyefficiently.In11,forexample,studentN76wasclearthat
thefocusoftheproblemwasnasalconsonants.
(11) N:Soifthereisnominimalpairthenwegotofigureouttheenvironmentthe
syllableisin.
R: Howwouldyoudothat?
N:Create a chart. So there are some words without [m], so I will probably
justignorethem.(N76)
Thisprotocolstatementsuggestsfurtherthatshehadadoptedastrategyoffocusingon
one nasal segment at a time, with the result that she was not distracted by irrelevant
data.Incontrast,withoutaclearrepresentationofthestructureoftheproblem,student
N53didnotknowwhatdatatoattendto.
(12) N:SoImightreadthewordsagainandjustlistdownonthepaperwherethe
separatephonemesare.IguessIwouldstartwiththefirstword[mut],so
therewouldbethreehere.
R: Areanyofthosenasalphonemes?
N:No,arewejustlookingfornasalphonemes?(N53)
Excerpt12suggeststhatthestudentwasattemptingtoexecutestep(c)fromtherecipe
in5,‘considerthedistribution’,buthadnotformedarepresentationofwhichsegments
wereunderconsideration,sospenttimeonirrelevantdetails.
Fromanexpertlinguist’spointofview,theenvironmentchartisatoolfororganizing
one’s knowledge about the problem.An environment chart can make it easier to ob-
servepatternsinthedistributionofaparticularsegment.Buttheprotocolssuggestthat
manystudentsperceivedtheenvironmentchartnotjustasausefulstepinthesolution
strategy, but as an end in itself.Avital challenge for instructors, therefore, is to make
clear the difference between tools for observing data on the one hand, and ways of
drawingconclusionsaboutdataontheother.
Incontrasttothenoviceswhojumpedintothesolutionstrategyrightaway,theex-
periencedstudentinvestedtimetorepresenttheproblemspacementally,asexample13
illustrates.
(13) E: Okay, so consonants involve [m], [ŋ], [n]. Mmm, okay. And voiceless
[m], voiceless [n] and voiceless [ŋ]. (whispering, reading instructions)
̥ ̥ ˳
TEACHINGLINGUISTICS e281
How many separate phonemes are there? How many of them are allo-
phones? What is your evidence for your conclusions? So let us put to-
getherourcommonones.
R: Whatareyoudoingrightnow?
E: I am reading through all of the words and categorizing the nasals that
match,that,atleastintheirsurfaceforms,looksimilartooneanothersoI
canseeiftherearecertainenvironmentsthatparticularnasalsarepopping
upin.(E53)
Student E53 first accessed her declarative knowledge about phonetics to determine
whichsegmentstoconsider,thenstartedtoidentifythedistributionofeach.Herstate-
ment‘letusputtogetherourcommonones’suggeststhatshehasamentalrepresenta-
tionofaphoneticsegmentasalinguisticobjectwithapredictablebehavior.Infact,this
allophonicvariation dis-
student’sconceptualrepresentationsofthenotionsof and
tribution
wererobustenoughthatshenotonlyhadnoneedtomakeanenvironment
chart to observe the relevant distributions, but also could not even remember what an
environmentchartiswhentheRApromptedherattheendofthethink-aloudsession.
These protocol statements from the outset of the problem-solving sessions show
hints of students’incipient linguistic concepts.They also show that beginner students
metacognitive
wereencounteringa thresholdconcept.Wismathandcolleaguescon-
cludedfromtheirstudyofundergraduates’metacognitiveproblem-solvingabilitiesthat
thethirdofthreethresholdconceptsinproblemsolving‘involvestherealizationofthe
importance of careful and complete modeling of a problem before one plunges into
tackling it’(Wismath et al. 2015:69). Since students in that study showed evidence of
reaching this threshold only toward the end of a course that concentrated entirely on
problem solving, it is not surprising that first-year linguistics students had not yet
achievedit.Instructorswhoofferarecipeforsolvingphonologyproblemsmightprof-
itablyincludeanearlystepintherecipesuchas‘thinkaboutwhatyoualreadyknow’to
remindstudentstoinvesttimeinrepresentingtheproblembeforestartingtosolveit.
Lookingforminimalpairs.
Thealgorithm(5)thatstudentspracticedinclassbe-
gins with a search for minimal pairs. Looking for minimal pairs is of course not logi-
callynecessaryinsolvingaphonologyproblem.Iadvisebeginnerstudentstostartwith
thisstepbecausewhentheyarescanningadataset,itcanbeeasiertolocateaminimal
pair than to notice patterns in the distributions of segments.About half of the novice
protocols(sixofeleven)indicatethatthestudentstartedtheproblembysearchingfor
minimalpairs,suchasin14.
(14) Firsttrytofindtheminimalpairs.Idon’tthinkthereareanyminimal—there
isonenearminimal—notreally.Trytofindthe[i]and[].Socircleallofthe
ɪ
[i] in the words. … I’m just trying to find and label them all first. I don’t
thinkthereisminimalpairs.SoIwilljuststartwithtryingtofindtheenvi-
ronmentstheyarein.(N76)
Inthesevenprotocolsthatdidnotmentionsearchingforminimalpairs,itispossible
thatparticipantseitherforgotthissteporwereconfusedaboutit,likestudentN78.
(15) I’m going to start with similar words and see if they contrast. So, similar
words,similarwords,comeon.UsuallywhenIlookforthesewords,Iusually
onlylookforthefirstletterandthenIgofromthere.Theyshowedmeinclass
tomakeatable,soI’mmakingatable.Ononesidethere’s[],thestressed[i]
ɪ
and[].ThenI’llputthewordsthataresimilarbesideeachother.(N78)
ɪ
e282 LANGUAGE,VOLUME92,NUMBER4(2016)
This portion of student N78’s transcript indicates some confusion among the steps in
therecipe.Althoughsheknewthatsearchingforminimalpairsandmakinganenviron-
mentchartwerebothstepsintherecipe,hermentalrepresentationwasnotclearonthe
functionsofthetwodifferentkindsofevidence.
Butformanyofthestudentswhoappearedtoskipthesearchforminimalpairs,itis
plausiblethattheycouldhaveactuallyperformedarapidscanand,notfindinganymin-
imal pairs, moved on immediately to observing phonetic environments without even
commentingaloudontheminimal-pairsearch.Inotherwords,notmentioningthemin-
imal-pair step of the recipe out loud does not necessarily indicate that they skipped
thatstep.
Asmentionedabove,theevidencewithintheAngasdatasetforphonemiccontrasts
between[m],[n],and[ŋ]issparse.Onenovicestudentobservedtheabsenceofmini-
malpairs.
(16) Okay,sothreenasalsareseparatephonemes.Buttherearenominimalpairs.
I’llneedminimalpairstosupportit.Buttheyareinthesameenvironment.
(N76)
Thisstudenthadaclearrepresentationofwhatkindofevidencewouldsupportacon-
clusionofseparatephonemes:shewouldhavepreferredthestrongevidenceofminimal
pairsbutwaswillingtodrawaconclusionbasedonthesounds’similardistributions.
Butmanystudents,eventheexperiencedstudent,simplyassumedthat[m],[n],and
[ŋ]arethreeseparatephonemeswithoutconsideringtheevidence.
(17) Inthiscase,therewouldbethreeseparatephonemes.Andum,howmanyof
them are allophones?We’ve got, okay, there are six different segments that
are appearing, three of which are voiced, three of which are voiceless.And
I’mprettysurethatyou’vegotatleastoneseparateallophoneperphoneme,
sothephonemeforexamplewouldbe[m],[n],and[ŋ],andthentheirallo-
phonesunderneaththem.(E53)
Even though the assumption appears to be correct in the case of this data set, making
the assumption in the absence of evidence is an error in linguistic thinking. No doubt
for some students this assumption arises from interference from their native language
English,where[m],[n],and[ŋ]areindeedthreeseparatephonemes.Studentsinthein-
troductorycourseoftenencountersuchinterferencewhentheyareexposedtodatafrom
languages whose phonology differs from English. For example, voicing allophony in
stops in Swampy Cree is often particularly challenging for students to recognize be-
cause the allophones are [p]-[b], [t]-[d], and [k]-[ ], each pair of which is contrastive
ɡ
inEnglish.
In addition to interference from English, the protocol statements also reveal other
flawsinstudents’conceptualrepresentationsofphonemiccontrast.TheIPAnotationit-
selfledsomestudentstomakeassumptionsaboutphonemesandallophones.Example
18suggeststhatstudentN76wasundertheimpressionthatallophonesareindicatedby
diacritics.
(18) [ŋ]happenshere,here,andhere.Buttheydon’thaveallophones.Howmany
of them are allophones, but they don’t have that circle there.4 There is no
characteristictodistinguishthose.SodoIstillneedtoconsiderthem?(N76)
Bythatkindofreasoning,segmentsrepresentedbyseparateIPAsymbolswouldauto-
maticallybeassumedtobeseparatephonemes.
4Thediacriticforvoicelessness.
TEACHINGLINGUISTICS e283
Student N84’s protocol statements reveal the weaknesses in her concept of phone-
miccontrast.Onesourceofconfusionwasbetweenphoneticdifferenceandphonemic
contrast.
(19) So then in this case, um, [n] is only found at the beginning but [m] is also
foundatthebeginningsothatdoesn’treallydistinguishthem.SoI’dsaythat
they’retwoseparatephonemesbecausethey’realsodistinguishedbythefact
that[n]isalveolarand[m]isbilabialsothey’redifferentinmorethantwo—
inmorethanonecriteria.(N84)
Theconfusionabouttheideaofcontrastemergedfromuncertaintyaboutthenatureof
minimal pairs. When Student N84 was struggling with the problem, the RAs offered
somecoachingprompts.
(20) R: OK,sowhat’saminimalpair?
N:Ifthey’reonlysimilarinonecriteria?
R: What’sacriteriainthiscase?
N:Idon’tknowifhastobethatthey’resimilarinmorethanonewayorif
they’redifferentinmorethanoneway,thenthey’reconsideredaminimal
pair.Iknowithassomethingtodo…somethingalongthoselines.
R: OK,canyougivemeanexampleofaminimalpair?
N:Sotherewaslikestunandspun.5AndIbelievethatwasconsideredamin-
imal pair. stun and spun—because—I’m checking to see if they’re like
voicedorvoiceless—[p],[t]—sothey’rebothvoiceless.Sothey’reonly
differentinthemannerofarticulation,Isuppose,solike[p]isbilabialand
[t]isalveolar,sothat’stheonlyway,like,theydiffer,sothen,that—that’s
whythey’reminimalpair. (N84)
N84’sresponsesrevealthatherconceptofaminimalpairwasapairofsegmentsthat
differminimally.Thismistakenimpressionmightarisefromparallelsbetweenthecon-
cepts:minimalpairsarepairsofwordsthatdifferinonlyonesegment,andphonology
problemsoftenfocusonpairsofsegmentsthatdifferinonlyonefeature.Ifastudent’s
understandingofminimalpairsisflawedinthisway,thenaconsequentconfusionbe-
tweenphonemesandallophonesisalmostinevitable.
Identifyingdistributions.
Thealgorithmin5thendirectsstudentstoconsiderthe
distributionoftherelevantsegments.FortheLebaneseArabicproblem,allofthestu-
dents,noviceandexperienced,arrivedatacorrectornearlycorrectdescriptionofthe
distributionofthesegments.Allrecognizedthat[i]appearedonlyword-finally.Theex-
periencedstudentandfiveofthesixnovicesrecognizedthedistributionof[]aselse-
ɪ
where relative to the word-final distribution of [i], though not always using the term
elsewhere. For example, participant N39 clearly understood that the underlying form
hasthemoregeneraldistribution,eventhoughshedidnotusethetermelsewhere.
(21) Becausethevowel,thetensevowel,appearstooccurattheendoftheword
alwaysanditdoesn’tshowanyotherplaceswhereitcouldoccur.Andthelax
vowel, there isn’t a situation where it occurs at the end of the word, so I’m
wondering if they pronounce it as the—they pronounce the lax vowel usu-
ally—I’mwonderingifit’susuallythelaxvowelandattheendoftheword
itjustturnsintoatensevowel.(N39)
5Fromarecentquiz.