Foreign accent strength and listener familiarity with an accent codetermine speed of perceptual ...

20 Pages·2013·0.49 MB·English

AttenPerceptPsychophys(2013)75:537–556
DOI10.3758/s13414-012-0404-y
Foreign accent strength and listener familiarity with an accent
codetermine speed of perceptual adaptation
MarijtJ. Witteman&AndreaWeber&
James M.McQueen
Publishedonline:2March2013
#PsychonomicSociety,Inc.2013
Abstract We investigated how the strength of a foreign accented speaker. Listeners were primed by the strongly
accent and varying types of experience with foreign- accented words, but again only in the experiment’s second
accented speech influence the recognition of accented half.Together,theseresultsshowthatadaptationtoforeign-
words.InExperiment1,nativeDutchlistenerswithlimited accentedspeechisrapidbutdependsonaccentstrengthand
or extensive prior experience with German-accented Dutch on listener familiarity with those stronglyaccented words.
completedacross-modalprimingexperimentwithstrongly,
medium,andweaklyaccentedwords.Participantswithlim- Keywords Foreign-accentedspeech .German-accented
ited experience were primed by the medium and weakly Dutch .Cross-modalpriming .Short-andlong-term
accented words, but not by the strongly accented words. experience .Perceptuallearning
Participants with extensive experience were primed by all
accent types. In Experiments 2 and 3, Dutch listeners with Itisestimatedthatmorethanhalfoftheworld’spopulation
limitedexperiencelistenedtoashortstorybeforedoingthe
speaksatleasttwolanguages(Grosjean,2010),withnumb-
cross-modal priming task. In Experiment 2, the story was
ers steadily growing. This multilingualism means that lis-
spoken by the priming task speaker and either contained
tening to foreign-accented speech has become a standard
strongly accented words or did not. Strongly accented ex-
listeningsituationinmetropolitanareas.Inforeign-accented
posureledtoimmediateprimingbynovelstronglyaccented
speech, native listeners are confronted with pronunciations
words, while exposure to the speaker without strongly
that deviate from their language standards (e.g., Dutch
accented tokens led to priming only in the experiment’s
speakers pronouncing kettle instead of cattle or, for native
second half. In Experiment 3, listeners listened to the story
Japanesespeakers,somethingsimilartofliedliceinsteadof
withstronglyaccentedwordsspokenbyadifferentGerman-
fried rice). How then do native listeners cope with foreign-
accented speech? Is understanding of foreign-accented
speech hinderedonlybylargedeviationsfromtheintended
: :
M.J.Witteman(*) A.Weber J.M.McQueen pronunciation?Caninexperiencedlistenersadaptquicklyto
MaxPlanckInstituteforPsycholinguistics,PObox310,
anewspeaker?Furthermore,afteradaptationtoonenonna-
6500AH,Nijmegen,TheNetherlands
tivespeaker,isitthenpossibletounderstandanotherspeak-
e-mail:[email protected]
er with the same accent? The present study addresses these
: :
M.J.Witteman A.Weber J.M.McQueen questions.
RadboudUniversityNijmegen,Nijmegen,TheNetherlands
:
A.Weber J.M.McQueen
DondersInstituteforBrain,CognitionandBehaviour, Variationin native speech
RadboudUniversity,Nijmegen,TheNetherlands
Although foreign accents add variation to speech, native
J.M.McQueen
speech contains considerable variability too. Most research
BehaviouralScienceInstitute,RadboudUniversity,
Nijmegen,TheNetherlands about variation in speech in fact stems from the first
538 AttenPerceptPsychophys(2013)75:537–556
language (L1) domain. Research on L1 speech shows that newitemswerepresented.Moreover,adaptationprovedtobe
evenwhenonephonemeinawordischangedarbitrarily,the extremely rapid; just a few trials were necessary to get a
word can still be recognized (e.g., Connine, Blasko, & (limited)effectofgeneralization.
Titone,1993;Marslen-Wilson,1993),atleastaslongasthis Processing of words is also affected by the frequency of
change does not create a new word (Marslen-Wilson & thevariant.Wordsthatarefrequentlypronouncedwithouta
Zwitserlood, 1989). However, the larger the phonemic de- schwa (e.g., corporateas corp’rate) were more likelyto be
viationinanonwordisfromthestandardpronunciationofa judged as two-syllable words than were words with a low
word, the harder it is to recognize the word correctly schwa-deletion rate (Connine et al., 1993; Connine,
(Connine et al., 1993). When words are changed in ways Ranbom,&Patterson,2008).Thisinfluenceofthedistribu-
typical for natural speech, as in greem bench, where the tion of variant representations has also been found in other
word green is assimilated to what sounds like greem be- languages, such as Dutch (Mitterer & Ernestus, 2006).
cause the following context warrants nasal place assimila- Experiments on the nasal flap in English (found in gentle
tion,thisdoesnotpreventrecognitionoftheintendedword inAmericanEnglish)havealsoshownthatlexicalrepresen-
(e.g., Gaskell & Marslen-Wilson, 1996, 1998; Gow, 2002; tationsarestrongerwhenpeoplehavemoreexperiencewith
Mitterer & Blomert, 2003). This effect does not occur, this phenomenon (Ranbom & Connine, 2007). Reductions
however,whenthefollowingcontextdoesnotlicenseplace also form an interesting perspective on frequency of occur-
assimilation (Gaskell & Marslen-Wilson, 1996, 1998; rence. Word-final /t/, for example, can be realized with an
Mitterer & Blomert, 2003). Reductions also form an inter- alveolar closure and an audible release or as a glottal stop
estingcaseforlookingatdeviation.Wordsthatarefrequent- without a release. English listeners were able to recognize
lyreducedinconversationalspeechcanberecognizedeasily both forms equally quickly in a semantic priming task;
as their intended targets; hearing posman, for example, will however, when there was a delay added to the task, partic-
facilitate recognition of unreduced postman (Ernestus, ipantsdidnoticeablybetterwiththetypicalvariant(Sumner
Baayen, & Schreuder, 2002; Mitterer & Ernestus, 2006). & Samuel, 2005). Degree of familiarity with foreign-
Together,thesefindingsthusshowthatalthoughtheamount accented pronunciations should thus influence how listeners
of deviation is important in recognizing word forms, the processthem.Inthepresentstudy,wetesteddifferenttypesof
context in which the variation occurs has to be appropriate familiarity effect: We first looked at the role of long-term
in order for successful recognition to occur. A foreign- experiencewithanaccentoncomprehensionandtheninvesti-
accented speaker provides a natural context for deviant pro- gatedhowrapidlycomprehensionofanunfamiliaraccentcould
nunciationsthatareconsistentwiththataccentbut,typically, improvewhenparticipantswerebrieflyexposedtothataccent
notwiththetargetlanguage. immediatelybeforetesting.Wethusaskedwhattypeofexpe-
rienceisneededtoprocessforeign-accentedspeechcorrectly.
Sumner and Samuel (2009) looked at the role of long-
Frequencyand familiarity effects term experience by studying dialectal variation (New York
City Englishvs.General AmericanEnglish) and found that
Frequency of occurrence is another important factor for speakersofaNewYorkCitydialectthatdropsword-final–r
recognition of deviant word forms. That is, variant forms (turningbak[ ]‘baker’intobak[ ])couldinstantlyinterpret
are more likely to be recognized correctly when they are these dialectal forms as the intended word, whereas non-
presentedinaphonologicalcontextinwhichtheyfrequently dialectal speakers (speakers of General American) did not
occur(suchascertainreductions,includingDutch['dam]for showsuchaneffect.Whentestedagainafterashort(20-to
/'darɔm/ [therefore] [e.g., Ernestus et al., 2002; Mitterer & 30-min) time lag, General American listeners had more
Ernestus, 2006], medial t-deletion, such as 'senner' or 'sen- trouble recognizing the dialectal forms than the standard
nah' for English 'center' [Pitt, 2009], and vowel raising forms, whereas for the New York City dialectal listeners,
[Dahan, Drucker, & Scarborough, 2008]). For example, there was no difference. Thus, listeners who are familiar
Pitt found that participants judged /t/-deleted variants as (passively or actively) with a dialect are apparently more
words only if the phonological environment in which the flexible in form processing than inexperienced listeners are:
/t/-deletion occurred was common in production. A similar Experiencedlistenerscandealwithmorevariation,relativeto
effectwasfound byDahanetal.,whomanipulatedEnglish thestandardpronunciation,whenlisteningtodialectalspeech.
words ending in /g/ and /k/ such that they would contain AsimilarresultwasfoundwithBritishEnglishlistenerswho
either a raised vowel (like the [ ] in bag) or an unraised had moved to the United States; they learned to interpret
vowel (similar to the [æ] in back). Half of the words had correctlythemedialflap(ɾ)in“todal”(/toɾal/)as/t/(thereby
contextually appropriate vowel raising, and the other half recognizing the intended word total; Scott & Cutler, 1984).
didnot.Participantslearnedtounderstandtheintendedword These listeners seemedtohaveadapted their perceptualsys-
andevenlearnedtoexpecttheincorrectpronunciationswhen temtoAmericanEnglishstandards.
AttenPerceptPsychophys(2013)75:537–556 539
Effects ofaccentstrengthin foreign-accented speech? Familiarity with German-accented Dutch
The present study focuses on foreign-accented speech As has already been noted, we were also interested in the
rather than dialectal variations. In foreign-accented effects of familiarity with a naturally occurring accent. In
speech, speakers often replace target language sounds particular,we tested whether Dutchlistenerswhoareeither
with native language sounds when the target speech familiarorunfamiliarwithGerman-accentedDutchareable
sounds are not found in the speaker’s native language to correctly interpret German-accented Dutch words.
(e.g., cattle pronounced as kettle by Dutch speakers; Research with second-language (L2) listeners suggests that
Broersma & Cutler, 2011). When speech sounds are recognition of familiar variant forms is possible (Weber,
shared between the two languages, however, substitu- Broersma, & Aoyagi, 2011). In an experiment in which
tions that involve a different category are usually not Dutch and Japanese participants listened to either Dutch-
observed. The former types of alteration are often per- accented English or Japanese-accented English, the L2 lis-
ceived as stronger accent markers than the latter. Thus, teners could recognize accented words easily when they
even within one language combination, there are words were produced in their own accent (e.g., Dutch listeners
that can be strongly affected by foreign accents and could recognize Dutch-accented English words easily, and
words that are affected to a smaller extent. Here, we Japanese listeners Japanese-accented English words). But
contrast the recognition of more strongly accented areparticipantsalsoabletoadapttoforeign-accentedspeech
words with that of less strongly accented words. when the target language is their native language? In that
Mostresearchonforeign-accentedspeechhasfocusedon case, listeners are constantly exposed to the native pronun-
its intelligibility (e.g., Bradlow & Bent, 2008; Derwing & ciationsfromtheirfellowcountrymen.NativeDutchspeak-
Munro,1997;Munro &Derwing,1995a).Ingeneral,intel- ers,forinstance,willusuallyhavefarmoreexperiencewith
ligibility increases as the strength of accent of a speaker nativeDutchthanwithGerman-accentedDutch.Cannative
decreases (Bent & Bradlow, 2003), and native listeners are listeners therefore easily understand only weakly accented
known to benefit from more exposure to improve their words,anddothey,inordertounderstandstronglyaccented
understanding of low-intelligibility speakers, whereas this words, need to attain a certain level of familiarity with the
isnotnecessaryforhighlyintelligiblespeakers(Bradlow& accent first?
Bent, 2008).Exposurefrommultiplespeakersisalsobene- Dutchlistenersparticipatedinthreecross-modalpriming
ficial for understanding foreign-accented speech (Sidaras, experiments with Dutch as the target language, although
Alexander,&Nygaard,2009).NativeEnglishlistenerswere spoken with a German accent. German-accented Dutch
familiarized with multiple speakers of Spanish-accented was chosen because native Dutch speakers are known to
English orwithnativeEnglishcontrol speakers. Participants vary in how familiar they are with the accent. There are a
were able to attain higher transcription accuracy on substantial number of German students in the Netherlands,
novel words and utterances after a familiarization phase but they tend to study at Dutch universities close to the
with multiple speakers, as compared with no exposure German border. Fewer German students are found in the
at all. Furthermore, participants who received accented center of the Netherlands. So it is possible to find Dutch
familiarization were better at recognizing some of the listeners with either limited or extensive experience with
accented vowels than were participants with only native German-accented Dutch.
training. Which vowels participants were exposed to and A German accent in Dutch is particularly noticeable
how this affected learning of these vowels, however, when it comes to vowels, and these vowels produce an
was not manipulated systematically. Most of the studies excellent starting point for looking at effects of degree of
on intelligibility have used tasks in which participants accentedness.Wethereforechosewordswithtwoparticular
were required to make judgments about sentences that Dutch diphthongs. Although both Dutch and German are
were not controlled for specific accent markers. The Germaniclanguages,theirvowelsystemsdifferinanumber
present study investigated specific accent markers and ofways.Bothlanguageshavealargevowelinventory:Dutch
their effect on understanding accented speech. We ex- has13monophthongsand3diphthongs(Gussenhoven,1999);
amined the recognition of individual words that vary in German has 12 monophthongs and 3 diphthongs (Kohler,
perceived accentedness. The difference in perceived 1999). While there is some overlap between the mono-
accentedness is mainly driven by vowels that deviate phthongs,thediphthongsvarymoreacrossthetwolanguages.
to a larger or smaller extent from the standard form. Dutch has the three diphthongs / /, /œy/, and /ʌu/, while
This perspective allowed us to investigate in greater German has /a/, / /, and / /. The two Dutch diphthongs
detail what makes foreign-accented speech hard to un- thatwerethefocusofthisstudy,[œy]and[ ],arethusnotpart
derstand and whether all types of variation lead to oftheGermanvowelinventory.Bothdiphthongsaredifficult
processing difficulty. for many learners of Dutch (Doeleman, 1998), but in
540 AttenPerceptPsychophys(2013)75:537–556
particular, /œy/ is a rare sound across languages and poses are known to be shorter when the auditory prime and the
great difficulties for L2 learners. The Dutch [œy]-vowel was visually presented target word are identical than when the
replaced with theGerman [ ] by the speakerof this experi- auditory prime is unrelated (see, e.g., Clarke & Garrett,
ment,andtheDutch[ ]wasreplacedwiththeGerman[a]. 2004; Marslen-Wilson, Nix, & Gaskell, 1995; McQueen,
Acoustically, the trajectories of [ ] and [œy] deviate more Cutler, & Norris, 2006). Phonologically similar but non-
thanthoseof[ ]and[a ].Thiswasconfirmedforthespeaker identical auditory primes generally do not produce signifi-
of our experiment with acoustical analyses (see the Method cant facilitatory priming and, sometimes, even produce
section for Experiment 1). German learners of Dutch are inhibitory priming (e.g., Van Alphen & McQueen, 2006).
usually well aware of the large deviation between their pro- We will therefore take statistically significant facilitatory
nunciationandtheintendedDutch[œy]-vowel,whiletheyare priming as our measure of successful online word recogni-
often oblivious to the smaller deviation between their [a] tion (Marslen-Wilson, Moss, & van Halen, 1996; Marslen-
pronunciation and the intended Dutch [ ]. In this study, Wilson & Zwitserlood, 1989) and, hence, of successful
Dutch words with [œy] were considered to be strongly adaptation to the accent. Facilitatory priming of responses
accented, and Dutch words with [ ] as medium accented. tovisualtargetwordsafterauditoryprimesproducedwitha
There was a third set of words without any segmental sub- foreign accent will thus be taken to suggest that listeners
stitutions. These items contained only phonemes shared recognized the accented primes online as being the same
between Dutch and German but were, nonetheless, spoken Dutch wordsas thevisualtargets.
by the same nonnative speaker. These words were there- Weexpectedthatrecognitionofthevariantforms would
fore considered to be weakly accented. The three different be more successful when the deviation from the standard
strengths of perceived accentedness were confirmed in a wassmaller(e.g.,thatwecouldobservesignificantpriming
rating study (see the Method section for Experiment 1). for the weakly accented words, but not for the strongly
In summary, the present study had three goals. First, in accented words) and that the more experience people had
Experiment 1, we investigated whether foreign-accented hadwithanaccent,theeasieritwouldbeforthemtoadapt
speech (in this study, German-accented Dutch) can be un- to it. Adaptation was measured separately for the first and
derstood by native (Dutch) listeners with limited previous second halves of the experiment. It was possible that less
exposure and contrasted their results with those of native experienced listeners would not show priming in the first
listeners who were already highly familiar with the accent. half but would in the second half, after having had time to
Second, we asked whether effects of familiarity depend on adapt to the accent. Given previous findings with native
how strongly accented the stimulus words were. Third, in speech (e.g., Marslen-Wilson & Zwitserlood, 1989), it was
two subsequent experiments, we examined how short-term predicted that Dutch listeners could interpret the weakly
trainingonanaccentinfluenceswordrecognition,againasa accentedwordscorrectly,regardless oftheirpreviousexpe-
function of strength of accent (Experiment 2), but also as a rience with the German accent. Furthermore, we expected
function of speaker (Experiment 3). The results of these thatparticipantswithlimitedpreviousexposuretoGerman-
experiments will be related to models of spoken-word rec- accented Dutch would have trouble understanding medium
ognition and the accounts they offer for how listeners cope and strongly accented words, whereas participants with
withpronunciationvariation—thosebasedonrepresentation extensive experience with the accent would have no prob-
(e.g.,Goldinger,1998;Johnson,2006;Pierrehumbert,2001; lems understanding such words.
Ranbom & Connine, 2007) and those based on processing
(e.g.,Gaskell&Marslen-Wilson,1998;Lotto&Holt,2006; Method
Mitterer, Csépe, Honbolygo, & Blomert, 2006).
Participants
Experiment 1 Twogroupsofparticipantsweretested.Thelimited-experience
group(n023,19females,meanage020.41years),allnative
Experiment1wasdesignedtotestwhetherfamiliaritywitha speakersofDutch,wastestedinUtrecht,acityinthemiddleof
foreign accent influences adaptation to that accent, as mea- the Netherlands. These participants were recruited from the
sured in terms of ease of word recognition. We compared UtrechtUniversityparticipantpool;thevastmajoritystudiedat
Dutch listeners with limited experience with German- UtrechtUniversity.Theextensive-experiencegroupalsocon-
accented Dutch with listeners with extensive long-term ex- sistedof21nativespeakersofDutch(19females,meanage0
perience with German-accented Dutch. Participants first 22.97years)andwastestedinNijmegen,acityintheeastof
listened toGerman-accentedDutch primes andthen judged the Netherlands, near the German border. These participants
whether targetwordsthatappearedonascreen wereDutch were recruited from the MPI participant pool; the majority
words or nonwords. Reaction times (RTs) to a target word studiedattheRadboudUniversityNijmegen.
AttenPerceptPsychophys(2013)75:537–556 541
There are many more German students enrolled at the words (38 nouns, 7 adjectives, 2 adverbs, and 1 pronoun),
Radboud University Nijmegen (approximately 8 % of the and their corresponding primes were either the German-
students in 2011 were German) than at Utrecht University accentedvariantsofthetargets(identicalprimes)orphono-
(approximately 1 % German students in 2011). As such, it logicallyandsemanticallyunrelatedDutchwords(unrelated
could be expected that Dutch students in Nijmegen are, in primes). The 48 identical primes comprised 12 strongly
general, more frequently exposed to German-accented accentedwords,12mediumaccentedwords,and24weakly
Dutch than students in Utrecht are. But students in Utrecht accentedwords.Strongly accentedprimeswerewordswith
can happen to have German friends or family with whom theDutchvowel[œy]asinhuis,‘house.’Thisdiphthongis
they communicate in Dutch, and students in Nijmegen can not part of the German phoneme inventory (see, e.g.,
happen to major in a subject where only a few German Kohler, 1999), and German learners of Dutch mostly sub-
students are enrolled. In order to control more closely for stituteitwith [ ],aGerman diphthongthatisperceptually
the amount of prior experience with German-accented and acoustically quite different from Dutch [œy] (Dutch
Dutch, a language history questionnaire was administered. [œy] starts front-central, half open, and ends front-central,
One of the questions asked how often participants heard nearclose;German[ ]startsback-central,halfclosed,and
German-accented Dutch (possible answers: never, less than ends front-central, near open).
once a week, once a week, and multiple times a week). MediumaccentedprimewordscontainedtheDutchdiph-
Another question asked from how many speakers partici- thong[ ]asinlijst,‘list’;thisdiphthongalsodoesnotexist
pantsheardGerman-accentedDutch(possibleanswers:0–1, in German (Kohler, 1999) and is usually substituted with
2–5,6–10,andmorethan10).OnlyUtrecht-basedstudents German [a] by German speakers of Dutch. The diphthong
who reported hearing German-accented Dutch less than [a ] is not present in the Dutch phonemic inventory
once a week from fewer than two speakers were included (Gussenhoven, 1999) but is phonetically relatively similar
in the limited-experience group, and only Nijmegen-based totheDutchdiphthong[ ].German[aɪ]beginscentral,half
participants who reported hearing German-accented Dutch open, and ends front, close-mid; Dutch [ ] begins front,
multiple times a week from more than two speakers were open-mid,and endsfront, close-mid.Bothvowelsthus end
included in the extensive-experience group. Because the in approximately the same place. We ensured that the
questionnaire was administered after the main experiment remaining sounds (consonants and other vowels) of the
(toavoidastrongfocusonGerman),anumberofadditional strongly and medium accented words did not contain other
participants were tested but were notincluded in theanaly- obvious segmental substitutions by using only phonemes
sis,becausetheydidnotmeetthesecriteria(14participants that were shared between the languages. Thus, except for
inUtrecht, 10 participants in Nijmegen). the investigated diphthongs, all sounds were part of the
InadditiontotheirexposuretoGerman-accentedDutch, Dutch and German phonemeinventories.
we also asked participants about their knowledge of To ensure that the experimental words could not be
German. In The Netherlands, all students in upper educa- interpreted as other existing Dutch words, we asked 10
tional levels have to take German language courses for at additional native Dutch participants who did not hear
least 3 years and will, thus, have some knowledge of German-accentedDutchmultipletimesaweektotranscribe
German. Since the educational programs for German are all 48 experimental words. Incorrect transcriptions were
very similar across the country, the knowledge acquired in given, by a maximum of 4 participants, to only four words
schoolshouldbecomparableforourlistenergroups.Indeed, (two strongly accented, one medium accented, and one
none of the participants in Experiment 1 studied German, weakly accented). The incorrect transcriptions included
none reported being fluent in German, and German was one existing Dutch word (one word, and by only 1 partici-
always reported to be either their second or the third non- pant),loan words (two words), andaname (oneword).All
native language. other words were transcribed by all participants as the
Allparticipantsvolunteeredandwerepaidasmallfeefor intendedwords.
participating. None reported a hearing disorder, and all had The weakly accented words contained only vowels and
normalorcorrected-to-normalvision.Theyallreportedthat consonants that are present in both the Dutch and German
English was the first nonnative language they had learned, phonemicinventory,suchas[ ]and[ɛ]vowelsand[m],[ŋ],
usuallystarting inschool around the age of10. and [b]. We thus minimized segmental variation and
strength of perceived accentedness. The same subset of
Materials additional vowels and consonants that was used for the
strongly and medium accented words was used for the
Therewere48criticalitemsand96fillers.Eachcriticalitem weaklyaccentedwords.Totheextentthatitcanbeexpected
wasacombinationoftwoauditoryprimewordsandavisual that these sounds contribute to the perceived accentedness
target word. The targets were Dutch mono- or bisyllabic of words, they should do so to a comparable extent for all
542 AttenPerceptPsychophys(2013)75:537–556
three accent types. The 48 unrelated primes matched, over- and German vowels separately in an hVba-context. This
all,innumberofphonemeswiththeircorrespondingtargets contextwaschosenbecauseitminimizesinfluencesofother
(e.g., ‘ketting’ [chain] and ‘prikkel’ [incentive]), and the segments on the vowels (Jenkins et al., 1997). All words
overall lexical lemma frequency of unrelated primes was were recorded at least twice per vowel in the speaker’s
notdifferentfromthefrequencyofthetargets(logfrequen- natural accent, in clear citation, from correct Dutch or
cy taken from the CELEX database [Baayen, Piepenbrock, German spelling (e.g., the investigated Dutch vowels were
&VanRijn,1993]),t(48)00.082,p0.935.Foracomplete written as huiba and hijba, the German vowels as Heuba
list ofcritical items, see Appendix 1. and Heiba). These recordings were made in the same ses-
Of the 96 filler items, 24 had a Dutch word as the visual sion as the recordings of the auditory primes. The best two
target.Theremaining72fillersallhadanonwordastheirvisual tokenspervowelwereselectedbyanativespeakerofDutch
target. Eighteen of these items contained [ ] or [a] in the (the first author). For the critical German and Dutch diph-
prime,sothatnotevery[ ]or[a]primewouldpredictayes thongs,wemeasuredthefirstandsecondformantsatthe25
response. Therefore, the overall ratio of words and nonwords and 75 percent points of the vowels. Figure 1 plots the
forthevisualtargetswas1:1,resultingin50%yesresponses average values for the first two formants for the Dutch
for errorless participants. Half of these 72 filler items were diphthongs[œy]and[ ],aswellastheGermandiphthongs
preceded by nonword primes; the others were preceded by [ ]and [a ] as produced by the speaker of the experiment,
existing Dutch word primes. Both the word and nonword as well as the diphthongs [œy] and [ ] spoken by Dutch
auditory primes could containthe [ ] and [a] vowels, again nativespeakers(datatakenfromAdank,VanHout,&Smits,
to ensure that participants could not form a response strategy 2004).
basedonthepresenceofthesevowelsintheitems. When looking at the Dutch / / in Fig. 1, it can be seen
thatourspeaker’sF1valueishigherthantheaverageDutch
Speaker selection speaker’s F1, but the trajectories are quite similar. In com-
parison, our speaker’s F2 is higher than the average Dutch
SevennativespeakersofGermanwererecordedwhileread- speaker’sF2forDutch/œy/,andthetrajectoriesarefurther-
ing a short Dutch text. Speakers differed in their level of more quite different. As can also be seen in Fig. 1, our
proficiencyandtimespentintheNetherlands.Theserecord- speaker’s Dutch / / and his German /a / are quite similar,
ings were made to find a learner of Dutch who would aswellashisDutch/œy/andGerman/ /.Thissupportsthe
produce the requested mispronunciations spontaneously notion that our speaker substituted the Dutch diphthongs
and consistently. The chosen speaker was a male native with existing German categories.
speaker of German, who grew up in Bavaria, in the south
of Germany. At the time of recording, he had lived in the Rating study
Netherlands for 2 years while studying in Dutch at the
Radboud University Nijmegen. The speaker was quite flu- To further ensure that the three types of identical primes
ent in Dutch and knew the meaning of almost all Dutch were indeed perceived as varying in accent strength, 20
words used inthe experiment.
The Dutch word and nonword primes were recorded in 00
4
pseudorandomized order from a list of items given in their
correct Dutch spelling. The speaker was not instructed to
0
changehispronunciation,soallmispronunciationsoccurred 50
naturally. The speaker produced the primes one by one,
separatedbyapause,inaclearcitationstyle,recordingeach
prime at least two times. The recordings were made in a F1600 œy
attempt at œy
sound-attenuated booth with a Sennheiser microphone and
were stored directly onto a computer at a sample rate of 00
7
44 kHz. Primes were excised from the recording using the
attempt at
speech editor Praat (Boersma & Weenink, 2009), and the
0
besttokenswereselectedbythefirstauthor,anativespeak- 80
er ofDutch. 2000 1800 1600 1400 1200 1000
F2
Acousticmeasurements Fig. 1 F1 and F2 formant values of the two critical diphthongs, as
pronounced by the experimental speaker (German-accented Dutch in
solid lines, native German in bold dashed lines), as compared with
Werecordedthecompletevowelspaceforourspeakerboth
Dutch diphthongs from the Adank et al. (2004) corpus (thin dashed
inDutchandinGermanbyhavinghimpronounceallDutch lines)
AttenPerceptPsychophys(2013)75:537–556 543
nativeDutchspeakerswhodidnotparticipateinthepriming primes. The visual targets stayed on the screen for
study rated the items. We used all strongly accented [œy]- 2,000ms,afterwhichthenexttrial started.Theexperiment
primes, all medium accented [ ]-primes, and half of the was created in Presentation (version 13, Neurobehavioural
weaklyaccentedprimes,sothattherewere12itemsofeach Systems Inc.) and was controlled with NESU hardware
type. We recorded two more sets of 36 items from two (NijmegenExperimentSet-Up).Afterthecross-modalprim-
additional speakers: a native speaker of Dutch and a native ing experiment, participants were asked to fill out the lan-
speaker of Italian with a very strong accent in Dutch. The guage history questionnaire.
reason for adding these two speakers was to add more
variation to the materials, thereby avoiding artificial infla- Results
tion of a perceived difference between prime types for the
Germanspeaker.Duringtheratingstudy,participantsheard Three items with weakly accented primes were discarded
one word at a time over closed headphones, immediately from the analysis because they had high error rates (more
followed by a rating scale where they could indicate how than10%).Apossiblereasonfortheseerrorratesisthatall
accented the word was on a scale ranging from 1 (not three of these targets had a very low lexical frequency (for
accented)to 10(very strongly accented). the exact items, see Appendix 1). These items were also
The data were analyzed with paired-samples t-tests that excluded from analyses of all other experiments described
indicatedthat,indeed,thestronglyaccented[œy]-items(M0 here.
7.98,SD01.18)wereratedasmoreaccentedthantheweakly The remaining cross-modal priming data were analyzed
accenteditems(M04.73,SD01.67),t(19)09.443,p<.001, with general linear model repeated measure analyses of
and the medium accented [ ]-items (M 0 7.01, SD 0 1.22), variance(ANOVAs)usinga3(accenttype:strong,medium,
t(19) 0 7.223, p < .001. Furthermore, the medium accented orweak)×2(priming:identicalvs.unrelated)×2(half:first
[ ]-items were rated as more accented than the weakly or second half of the experiment) design. All of these were
accented items, t(19) 0 7.576, p < .001. These results thus within–participants factors. We analyzed the results sepa-
confirmthepictureemergingfromtheacousticmeasures:The rately by participants (F ) and items (F ). In addition to
1 2
strongly accented words deviated most from the standard these analyses, we conducted planned comparisons using
pronunciations and were, indeed, rated as more strongly paired sample t-tests to look at the priming effects. These
accentedthanthemediumaccenteditems. were calculated separately by half and accent type.
Design and procedureof primingexperiment Limited-experience group
Participants were seated in a sound-attenuated booth and In addition to the three weakly accented items, a further
were informed that they would first hear a Dutch word or 1.8%oftrialsonwhichparticipantsmadeerrors,aswellas
nonword spoken by a German-accented speaker and then trials with RTs that deviated more than 2.5 SDs from the
seeaDutchwordornonwordonthescreen.Twoversionsof condition’soverallmean,werediscarded(together,<5%of
the cross-modal priming experiment were created, so that all trials). Errors were distributed evenly across the condi-
every participant saw each visual target only once. To con- tions (see Appendix 2) and, due to their low overall occur-
trol for effects of presentation order, each participant re- rence, will not beanalyzed statistically.
ceived a different pseudorandomized list. The first two As is shown in Fig. 2 (calculated priming effects) and
items of the experiment were always fillers, and there were Appendix 2 (mean RTs), participants with limited experi-
never two critical items ina row. ence were faster to respond to identical than to unrelated
The participants’ task was to decide as quickly and ac- trials, F (1, 22) 0 49.028, p < .001; F (1, 11) 0 86.190,
1 2
curately as possible whether the word presented on the p < .001. A main effect of accent type across partic-
screen was an existing Dutch word or not. Participants ipantsfurthermoreindicatedthatparticipantsreactedtotarget
responded by pushing one of two buttons on a button box wordswithdifferentspeeds,F (2,44)03.876,p0.028;F (2,
1 2
in front of them. Yes responses were always made with the 22) 0 1.100, p 0 .350. Participants also got faster, overall,
dominant hand, and RTs were measured from visual target during the course of the experiment, F (1, 22) 0 7.536,
1
onset. p 0 .019; F (1, 11) 0 7.267, p 0 .013.
2
Auditory primes were presented binaurally over closed The F analysis showed a significant interaction across
1
headphones at a comfortable listening level. Participants participants between accent type and priming, F (2, 44) 0
1
saw the visual targets on a computer screen situated about 4.580,p0.016; F (2,22)02.541,p0.102,reflectingthat
2
50 cm in front of them. Visual targets were presented in participantsdidnotshowequalprimingeffectsforallaccent
white lowercase 24-point Tahoma letters on a black back- types. Thiswas investigated further using planned pairwise
ground, 500 ms after the acoustic offset of the auditory comparisons(seeTable1).Theinteractionbetweenhalfand
544 AttenPerceptPsychophys(2013)75:537–556
200 F1(1, 20) 0 5.301, p 0 .033; F2(1, 11) 0 8.957, p 0 .014.
180 Strongly-accented items Across participants, there was a marginally significant
ms) 160 MWeeadkiulym-a-acccceenntetedd i tietemmss i3n.0te7r3a,cptio0n.0b5e8tw;Feen<1ac,creefnletcttyinpgethanatdpprirmiminignge,ffFec1t(s2d,i4ff0e)re0d
nrelated - identical; 11142080000 irpneraismPcihzliaensngfingoeerndfitffheipeccaatasincrcwcfeoe2i.nrsteetayccpohemsap.cTacrehinseotrnetysmpwaeinesrienepgaurisnaettdeelryato.ctTiloaonboslkeda2idtdntihose-t
u
g effect ( 6400 pplaarytiscitphaentsstaatnisdticiatelmasn.alPyasretsiciopfanthtse wpriitmhienxgteenfsfievcetseaxcpreorsis-
n
mi 20 enceshowedsignificantprimingfromthestartoftheexper-
pri
RT 0 iment for all accent types, thereby showing that they had
previouslyadaptedtotheaccentandwereabletoapplythis
-20
knowledge rapidlyto theexperimental speaker.
-40
First Second First Second First Second
Experiment half Discussion
Fig.2 Experiment1:Primingeffectsandconfidenceintervalsforpar-
Taken together, the results of Experiment 1 thus show that
ticipantswithlimitedexperiencebyexperimenthalvesandaccenttype
listeners with limited experience with German-accented
priming was not significant, F < 1, F < 1, nor was the Dutchandthosewithextensiveexperiencecanimmediately
1 2
three-wayinteraction,F <1,F <1.Thepairwisecompar- interpret the medium and weakly accented items correctly,
1 2
isonsshowthatparticipantshadnoproblemswithadapting butonly thelistenerswith extensiveexperienceshow facil-
to the weakly and medium accented items but could not itatory priming for strongly accented items. We also ana-
adapt to thestrongly accented items during theexperiment. lyzedbothlistenergroups(limitedexperienceandextensive
experience) in one overall repeated measures analysis with
Extensive-experience group group as an additional between-participants factor. These
analyses showed an effect of group marginally significant
Trials on which participants made errors (2.0 %) were acrossparticipants,F (1,43)03.737,p0.060;F (1,20)0
1 2
excluded from the analyses. In addition, we excluded trials 18.223,p<.001,withtheexperienced listenersresponding
on which the RTs deviated more than 2.5 SDs from the faster, overall, than the inexperienced listeners, but there
condition’s overall mean (together, <5 % of all trials). were no significant interactions between any of the factors.
Inspection of theerrorsshowed nospecific patterns. This could imply that the difference between the groups
As is shown in Fig. 3 (calculated priming effects) and with respect to processing strongly accented words was
Appendix2(meanRTs),participantswithextensiveexperi- possiblylesspronouncedthantheseparateanalysessuggest
encewithGerman-accentedDutchwerefastertorespondto orthatthiseffectdependsonthestrengthofelementsofthe
identicalthantounrelateditems,F (1,20)081.727,p<.001; accent. Moreover, it is possible that even Dutch listeners
1
F (1,11)0 29.828,p<.001,andrespondedsimilarlytothe with extensive experience with German-accented Dutch
2
threedifferentaccenttypes,F <1;F <1.Participantswere maystillencountersomedifficultiesunderstandingstrongly
1 2
faster in the second half of the experiment than in the first, accented words.
Table1 Plannedpairwisecomparisonsofprimingeffectsforallaccenttypesacrossparticipantsanditemsforparticipantswithlimitedexperience
withGerman-accentedDutch
Condition Half Participantanalysis Itemanalysis
df t p df t p
1 2
Stronglyaccenteditems First 1,22 1.233 .230 1,11 0.997 .340
Stronglyaccenteditems Second 1,22 1.399 .176 1,11 1.781 .103
Mediumaccenteditems First 1,22 4.026 .001 1,11 3.348 .007
Mediumaccenteditems Second 1,22 3.414 .002 1,11 3.546 .005
Weaklyaccenteditems First 1,22 4.803 .000 1,20 3.599 .002
Weaklyaccenteditems Second 1,22 4.947 .000 1,20 2.301 .032
AttenPerceptPsychophys(2013)75:537–556 545
200 Experiment 2
180 Strongly-accented items
Medium-accented items InExperiment2,weaskedforthefirsttimewhetherthereis
ms) 160 Weakly-accented items short-term adaptation to individual vowels in a real foreign
ntical; 140 accent (as opposed to an artificial accent). We again tested
nrelated - ide 112000 DaimcucmtecnehtdeidlaitsDetleuyntcebhre,sfbowuretitthhtheeylimwcreiotresesdn-moewxodpeaexlrpieponrsicemdeitnwogiateh4x-pmGeiernirmmsteoannrty-.
u
ng effect ( 8600 Tusheedsihnotrhtesptorirmyiwngasexsppeorkimenenbtyanthdewsoaumldetshpuesakfuenrcatisonthaast
mi additional exposure to that speaker. Two versions of the
T pri 40 story were recorded, one with 12 strongly accented items
R
20 (words containing the [œy]-vowel) not used in the main
experiment, and one without strongly accented items. The
0
goal of this experiment was to investigate whether a short
First Second First Second First Second
Experiment half period of familiarization with the speaker would be suffi-
cient to create adaptation to novel words. We expected to
Fig.3 Experiment1:Primingeffectsandconfidenceintervalsforpar-
replicate the findings of Experiment 1 with respect to the
ticipantswithextensiveexperiencebyexperimenthalvesandaccenttype
medium and weakly accented words (i.e., priming in both
Experiment 1 allowed us to study the role of long-term halvesoftheexperiment).Wealsoexpectedtofindpriming
exposure to German-accented Dutch. It showed not only for the strongly accented words when participants with
thatthedegreeofexperienceinfluenceshoweasilyalistener limited experience had been exposed to a story containing
canadapttoforeign-accentedspeech,butalsothatthiseffect strongly accented words (in contrast to Experiment 1) at
depends on the strength of the accent. In Experiment 2, we leastinthesecondhalfand,possibly,alreadyinthefirsthalf
wanted to shed more light on the role of short-term experi- oftheexperiment.Weexpectedparticipantswholistenedto
ence in word recognition and, inparticular, on the effect of theweaklyaccentedexposurealsotoshowadaptationtothe
short-term exposure on adaptation to strongly accented strongly accented items, but later or to a lesser extent than
words. Research on perceptual learning has shown that the listeners who were exposed to the story with strongly
word recognition process not only is sensitive to long-term accented words.
listeningexperience,butalsocanadaptafterashortamount
ofexposure.Thereissomeevidenceforrapidadaptationto Method
foreign-accented speech (see, e.g., Bradlow & Bent, 2008;
Clarke & Garrett, 2004). This is in line with perceptual Participants
learning research, with artificial accents or speech sounds
constructed to be ambiguous showing flexibility in online Fifty-seven participants took part in Experiment 2.
word recognition processes (e.g., Maye, Aslin, & Participants were randomly assigned to one of two story
Tanenhaus, 2008; McQueen et al., 2006), as well as in the exposure groups (group 1, n 0 19, 15 females, mean age 0
perceptionofphonemecategories(e.g.,Eisner&McQueen, 21.18 years; group 2, n 019, 16 females, mean age 0
2006; Norris, McQueen,& Cutler, 2003). 22.15 years). Participants from Utrecht were selected using
Table 2 Planned pairwise comparisons of priming effects for all accent types across participants and items for participants with extensive
experiencewithGerman-accentedDutch
Condition Half Participantanalysis Itemanalysis
df t p df t p
1 2
Stronglyaccenteditems First 1,20 3.009 .007 1,11 1.302 .220
Stronglyaccenteditems Second 1,20 4.855 .000 1,11 1.871 .099
Mediumaccenteditems First 1,20 3.562 .002 1,11 2.079 .032
Mediumaccenteditems Second 1,20 5.172 .000 1,11 3.642 .004
Weaklyaccenteditems First 1,20 6.012 .000 1,20 2.155 .016
Weaklyaccenteditems Second 1,20 3.263 .004 1,20 2.156 .050
546 AttenPerceptPsychophys(2013)75:537–556
the same criteria as in the limited-exposure group in instructions for the cross-modal priming experiment on the
Experiment 1. As before, therefore, we excluded additional screen. The cross-modal priming experiment was identical
participants who heard German-accented Dutch more than to the one used in Experiment 1. After the experiment,
once a week from more than two speakers. In total, we participants filled out the language history questionnaire
excluded 19 participants. As in Experiment 1, participants (identical to that inExperiment 1).
did not report a hearing disorder and had normal or
corrected-to-normal vision. All of them reported basic Results
knowledge of German, with German being their second or
third nonnative language (after English); none considered Limited-experience group–exposure with strongly accented
themselves fluent inGerman. words
Materials The statistical analyses were identical to those used in
Experiment1.Trialscontainingerrorswereexcluded(1.3%)
Two versions of a short story were created. We chose to fromthe analyses. We alsoexcluded trials onwhichtheRTs
focus on the effects of exposure to strongly accented deviatedmorethan2.5SDsfromthecondition’soverallmean
words, because learning for the medium and weakly (together, <5 % of all trials). Inspection of the errors (see
accented words was already at ceiling in Experiment Appendix2)againshowednospecificpatternsofpriming.
1. Both stories were recorded by the speaker from The priming effects for participants exposed to the story
Experiment 1. The story was based on the fairytale withstronglyaccentedwordsaredepictedinFig.4(meanRTs
“Jorinde and Joringel” by the Brothers Grimm. The are given in Appendix 2). Participants were faster overall in
two versions were identical, except for 12 words. One responding to identical trials, as compared with unrelated
version contained 12 strongly accented words with the trials,F (1,18)082.771,p<.001;F (1,11)0132.515,p<
1 2
[œy]-vowel (pronounced by the speaker as [ ]). None .001. The participant analysis showed that participants
of these words appeared in the main experiment. In the responded with different speeds to the three word types,
other version, these 12 words where replaced with F (2, 36) 0 3.788,p 0 .033; F < 1.Participantswerefaster
1 2
words without the vowel [œy] (e.g., duizend, ‘thousand’ in the second half of the experiment than in the first half,
replaced with honderd, ‘hundred’). The two stories thus althoughthiseffectwassignificantonlybyitems,F (1,18)0
1
both contained medium and weakly accented words. 3.638,p0.074;F (1,11)06.754,p0.025.Therewereno
2
Both versions were recorded from a script using correct significantinteractionsbetweenthefactors.
Dutch spelling. The speaker was not instructed to Planned pairwise comparisons were used to confirm the
change his pronunciation; again, all mispronunciations priming effects. Table 3 displays the priming effects across
occurred naturally. The speaker recorded one paragraph participants and items. Participants showed successful ad-
at a time. All paragraphs were recorded multiple times, aptationtoallaccenttypesinbothhalvesoftheexperiment.
and the best paragraphs were selected to create a story
that was spoken without hesitations and misreadings.
200 Strongly-accented items
That is, recordings were selected in which all words Medium-accented items
180 Weakly-accented items
were pronounced as intended, including the 12 strongly
accented words consistently pronounced with [ ]. ms) 160
Recordings were made in a sound-attenuated booth with ntical; 140
a Sennheiser microphone and were stored directly onto de
a computer at a sample rate of 44 kHz. nrelated - i 112000
u
Design and procedure ng effect ( 8600
Participants were seated in a sound-attenuated booth and mi
were informed that they would first listen to a Dutch story T pri 40
R
spoken by a German speaker and then perform a cross-
20
modal priming experiment. There was no additional task
0
when participants were listening to the story. Group 1 lis-
First Second First Second First Second
tenedtothestorywith12stronglyaccentedwords.Group2 Experiment half
listened to the story without strongly accented words.
Fig.4 Experiment2:Primingeffectsandconfidenceintervalsforthe
Participants were randomly assigned to a group.
limitedexperiencegroupwithexposuretostronglyaccenteditemsby
Immediately after the story ended, participants saw the experimenthalvesandaccenttype