Irrelevant speech, articulatory suppression, and ... · PDF fileA test of the phonological...

Psychonomic Bulletin amp Review2003 10 (2) 435-444

Immediate serial recall of verbal material is reducedsignificantly if participants are exposed to irrelevant speechwhile studying the list items This effect occurs despite thefact that participants are told they should ignore the irrel-evant material The irrelevant speech effect was first re-ported by Colle and Welsh (1976) with visual presentationof list items and has since been demonstrated in a largenumber of published studies (eg Ellermeier amp Zimmer1997 Jones 1994 Jones amp Macken 1995 Jones Mad-den amp Miles 1992 Salameacute amp Baddeley 1982 SurprenantLeCompte amp Neath 2000) The irrelevant speech effectalso occurs when the list items are presented auditorily(eg Hanley amp Broadbent 1987 Neath Surprenant ampLeCompte 1998)

Recently Neath (2000) has put forward a computationalmodel of serial recall known as the feature model that at-tempts to explain why irrelevant speech reduces recall per-formance One of Neathrsquos main goals was to provide a plau-sible explanation of the way in which irrelevant speechinteracts with three other variables that are also known toaffect serial recall articulatory suppression phonologicalsimilarity and word length Neathrsquos model differs in anumber of ways from Salameacute and Baddeleyrsquos (1982) ac-

count of the interrelationship between these variableswhich is based on the phonological loop component ofworking memory (eg Baddeley 1986) The present papertests the competing predictions that these two theoriesmake about the relationship between irrelevant speech andarticulatory suppression (Experiment 1) and about the re-lationship between irrelevant speech and phonologicalsimilarity (Experiment 2)

Phonological Loop Account of the IrrelevantSpeech Effect

According to Salameacute amp Baddeley (1982) the effects ofirrelevant speech are explicable in terms of the operationof the phonological loop (eg Baddeley 1986 19901992) Salameacute and Baddeley (1982) argued that irrelevantspeech interferes with the temporary storage of verbal ma-terial within a phonological input store of limited capac-ity The phonological store together with the process ofarticulatory rehearsal constitute what is known as thephonological loop Spoken material has obligatory accessto the phonological store whereas visually presented in-formation will enter the phonological store if it is articu-lated Figure 1 provides a flow diagram that illustrates someof the functions of the main components of the phonolog-ical loop model Consistent with this account Salameacute andBaddeley (1982) showed that when the list items were pre-sented visually the effects of irrelevant speech were re-moved by articulatory suppression Similar findings were

435 Copyright 2003 Psychonomic Society Inc

Correspondence should be addressed to J R Hanley Department ofPsychology University of Essex Wivenhoe Park Colchester CO4 3SQEngland (e-mail rhanleyessexacuk)

Irrelevant speech articulatory suppressionand phonological similarity

A test of the phonological loop modeland the feature model

J RICHARD HANLEY and EIRINI BAKOPOULOUUniversity of Essex Colchester England

Two experiments tested competing predictions about the nature of the irrelevant speech effect thatwere derived from Neathrsquos (2000) feature model and from Salameacute and Baddeleyrsquos (1982) phonologicalloop model The first experiment examined the combined effects of irrelevant speech and articulatorysuppression when target items were presented auditorily Contrary to the suggestions of Neath butconsistent with the phonological loop model the effects of articulatory suppression and irrelevantspeech were additive even when the irrelevant speech was presented during the retention interval Thesecond experiment examined the combined effects of irrelevant speech and phonological similaritywhen target items were presented visually Consistent with the phonological loop model the effects ofphonological similarity and irrelevant speech were additive when participants were specifically in-structed to use articulatoryphonological rehearsal to remember the list items The results thereforecontradicted Neathrsquos claim that irrelevant speech abolishes the phonological similarity effect when listitems are presented visually However the effect of phonological similarity was abolished in the irrel-evant speech conditions when no instructions were given concerning rehearsal It is argued that thephonological similarity effect disappears in some experiments because participants sometimes em-ploy a semantic rehearsal strategy consistent with the views of Salameacute and Baddeley (1986)

436 HANLEY AND BAKOPOULOU

reported by Hanley (1997) Salameacute and Baddeley (1982)claimed the reason for this interaction is that concurrentarticulation isolates visually presented material from thephonological store Although performance will be loweras a consequence it will not be further reduced by expo-sure to irrelevant speech

It follows from the phonological loop model that if thetarget material is presented auditorily then irrelevantspeech should reduce serial recall performance under ar-ticulatory suppression This is because the auditorily pre-sented list items will enter the phonological store auto-matically As a consequence list items will be subject tointerference from irrelevant speech even though concurrentarticulation is taking place This prediction was supportedby Hanley and Broadbent (1987) They showed additiveeffects of irrelevant speech and articulatory suppressionon the recall of auditorily presented lists of items in twoout of three experiments even when articulatory suppres-sion continued during presentation and recall We investi-gate this issue further in Experiment 1

The phonological loop model also provides an accountof the relationship between the effects of irrelevant speechand the effects of phonological similarity (Baddeley1966 1968 Conrad amp Hull 1964) on serial recall It isclaimed (eg Baddeley Lewis amp Vallar 1984) that therepresentations of phonologically similar items cause in-terference to one another within the phonological storeConsistent with this claim a number of studies haveshown that with visual presentation of list items the ef-fects of phonological similarity are also abolished by ar-ticulatory suppression (eg Baddeley et al 1984 Lon-goni Richardson amp Aiello 1993 Murray 1967 1968)Because both the irrelevant speech effect and the phonolog-ical similarity effect reflect the operation of the phono-logical store it follows from the phonological loop model

that the effects of phonological similarity and irrelevantspeech should be additive In other words phonologicallysimilar items should be recalled worse than phonologi-cally different items even in the presence of irrelevantspeech Although this prediction has been supported insome studies (Jones amp Macken 19951 Larsen Baddeleyamp Andrade 2000 Salameacute amp Baddeley 1986) the evi-dence is mixed Neither Colle and Welsh (1976) nor Sur-prenant Neath and LeCompte (1999) observed a phono-logical similarity effect in the presence of irrelevant speechWe investigate this issue further in Experiment 2

Feature Model Accountof the Irrelevant Speech Effect

Neathrsquos (2000) alternative model of the irrelevantspeech effect is based on Nairnersquos (1990) feature modelThe feature model assumes that each item from a list isrepresented in memory as a discrete set of features that areset to either 1 or 21 Every item is represented in both pri-mary memory and secondary memory Although sec-ondary memory traces are considered to be intact tracesin primary memory are subject to interference from lateritems The presentation of a new item that shares a featurevalue with one or more items from earlier in the list willhave the effect of altering the value of that feature from 1or ndash1 to zero in the memory traces of all the previouslypresented items This renders the feature useless for thepurposes of recalling the earlier items At recall eachtrace in primary memory is used to access the most simi-lar trace in secondary memory with the trace that hasgreatest feature overlap with the trace in primary memorybecoming selected for recall As features become lostfrom the trace of the item in primary memory so the prob-ability increases that an incorrect item will be recalled

The features that represent list items are either modality-independent or modality-dependent features The formerare activated regardless of modality of presentationwhereas different modality-dependent features are acti-vated when an item is presented visually from those thatare activated when it is presented auditorily It is assumedthat there are more modality-dependent features associ-ated with the auditory modality than with the visual modal-ity (hence the modality effect in serial recall)

The Effects of Irrelevant Speech andArticulatory Suppression

According to Neath (2000) irrelevant speech and artic-ulatory suppression reduce recall in broadly similar waysThe word that is repeated by the participant during articu-latory suppression causes additional interference to previ-ously presented list items Some of the modality-independentfeatures in primary memory are replaced by features ofthe word that is articulated a process referred to as featureadoption To simulate this effect half of the modality-independent features in each trace are set to a constantvalue (11) Feature adoption is also the cause of the in-terference that is brought about by irrelevant speechSome of the features of the words in the irrelevant speechstream replace features of the traces of the target items in

Figure 1 A representation of the phonological loop modeltaken from Gathercole and Baddeley (1993) Speech inputs (egirrelevant speech and auditorily presented list items) gain auto-matic entry to the phonological store Nonspeech inputs (eg vi-sually presented list items) require articulatory rehearsal inorder to access the phonological store and are therefore cut offfrom the phonological store by articulatory suppression Irrele-vant speech will interfere with the representations of list items ifand only if they are being held within the phonological store Theform in which visually presented words or letters are stored undersuppression is not specified by the model

IRRELEVANT SPEECH 437

primary memory This process is simulated by replacinghalf of the modality-independent features in each trace bya random combination of 11s and 21s2

On the basis of the claim that irrelevant speech and ar-ticulatory suppression exert similar effects on traces inprimary memory Neath (2000) argued that there will beno extra interference from irrelevant speech under condi-tions of articulatory suppression Here the predictions ofthe feature model are in conflict with the predictions of thephonological loop model The phonological loop model andthe feature model both predict that there will be no effectof irrelevant speech under conditions of articulatory sup-pression when list items are presented visually With audi-tory presentation of list items however the phonologicalloop model predicts additive effects of articulatory sup-pression and irrelevant speech whereas the feature modelpredicts a reduced effect of irrelevant speech when artic-ulation is suppressed These rival predictions were inves-tigated in Experiment 1

The Effects of Irrelevant Speechand Phonological Similarity

Under normal conditions in Neathrsquos (2000) simulationsa random sample of 50 of the features in any two traceswill be set to the same value The effects of phonologicalsimilarity are simulated by setting a consistent set of featuresto the same value in phonologically related items Thismeans that there will be a higher proportion of shared fea-tures in the traces of list items than would be the case inphonologically different lists By the time of recall there-fore virtually all of these feature values will have beenturned to zeros in the primary memory traces and perfor-mance will be relatively poor as a consequence With visualpresentation of list items the feature model predicts thatthere will be no effect of phonological similarity underconditions of irrelevant speech This is because recall of vi-sually presented material depends heavily on modality-independent features (there are limited modality-dependentfeatures associated with the visual modality) However ir-relevant speech has already interfered with the representa-tions of modality-independent features and phonologicalsimilarity will produce no further damage to them

Here again the predictions of the feature model and thephonological loop model are in conflict The phonologi-cal loop model predicts that irrelevant speech and phono-logical similarity should be additive for visually presentedlist items As we saw earlier previous research on this issuehas produced mixed results with only some of the pub-lished studies demonstrating additive effects of phono-logical similarity and irrelevant speech In Experiment 2we investigated possible reasons for these inconsistencies

EXPERIMENT 1

Neathrsquos (2000) feature model predicts that there shouldbe no effect of irrelevant speech under articulatory sup-pression when list items are presented auditorily (see theright panel of Simulation 6 Neath 2000 p 418) The onlypreviously published investigation of this issue was pro-

vided by Hanley and Broadbent (1987) In their first ex-periment they found that articulatory suppression re-moved the irrelevant speech effect However there weresome reasons why the sensitivity of this experiment mayhave been limited First different participants were usedin the speech and quiet conditions Second list length waslong (nine items per list) and it has been claimed that par-ticipants might stop using the phonological loop alto-gether when conditions are very difficult (Baddeley 2000)In their final experiment Hanley and Broadbent employeda within-subjects experiment and attempted to make thetask easier by reducing list length to six items per listUnder these circumstances there were additive effects ofirrelevant speech and articulatory suppression consistentwith the phonological loop model Neath suggested thatthese results might have come about because the irrele-vant speech masked the perception of the auditorily pre-sented target items This masking would have caused anadditional detrimental effect on recall under conditions ofarticulatory suppression Hanley and Broadbent (p 291) ar-gued against such an explanation by demonstrating unim-paired perception of auditorily presented target itemswhen accompanied by irrelevant speech Experiment 1sought to investigate this issue further by presenting the ir-relevant speech during the retention interval between pre-sentation and recall instead of during input

In Experiment 1 therefore a technique used previouslyby Miles Jones and Madden (1991) was employed Usingvisual presentation of target material Miles et al showedthat the effects of irrelevant speech were as strong when itwas presented during the retention interval as when it waspresented at the same time as the list items In Experiment 1we used Miles et alrsquos procedure with auditory presenta-tion of list items Half of the participants were exposed toirrelevant speech during the retention interval and halfwere exposed to irrelevant speech during list presentationIf some of the effects of irrelevant speech on auditorilypresented list items come about because of perceptualmasking at encoding then these effects should be signif-icantly reduced if the irrelevant speech is presented duringthe retention interval only If there is an effect of irrele-vant speech under articulatory suppression simply be-cause the irrelevant speech masks the target items thenthere should be additive effects of irrelevant speech andarticulatory suppression when irrelevant speech is pre-sented during list presentation When irrelevant speech ispresented during the retention interval there should be aninteraction between irrelevant speech and articulatorysuppression such that there is no effect of irrelevant speechin the articulatory suppression conditions The phonolog-ical loop model on the other hand predicts that the irrel-evant speech should exert similar effects on recall whenarticulation is suppressed regardless of whether the speechoccurs during list presentation or the retention interval

MethodParticipants Forty students at Essex University volunteered to

take part in the experiment and were tested individually Twenty par-ticipants were assigned at random to the two experimental groups


Design The experiment employed a mixed factorial design Lo-cation of irrelevant speech (during list presentation or during the re-tention interval) was manipulated between subjects Irrelevantspeech (present vs absent) and articulatory suppression (present vsabsent) were manipulated within subjects Each participant there-fore took part in four different conditions control articulatory sup-pression alone irrelevant speech alone and articulatory suppressionplus irrelevant speech The order in which participants performedthese conditions was counterbalanced by the use of a Latin square

Materials All lists contained six target items The lists com-prised randomly ordered sequences of the digits 1ndash9 sampled withreplacement A total of 60 such lists were prepared with 15 lists ineach of the four conditions The first three lists in each block weretreated as practice lists Materials were presented by an Apple Mac-intosh using the Quicktime application The Final Cut-Pro applica-tion was used to edit the tapes

Procedure Each participant wore stereo headphones for the du-ration of the experiment A 1-sec tone presented in the right head-phone alerted participants to the start of each list One second afterthe tone they heard the target items through the right headphonespoken by a female voice at a rate of one digit per second Each par-ticipant listened to 60 lists of items separated into four blocks con-taining 15 lists each with a break between blocks while the new setof instructions was presented For two blocks once in the presenceof irrelevant speech and once without the participant had to repeatthe word and at a rate of approximately two repetitions per secondduring presentation retention interval and recall of the target stim-uli Participantsrsquo articulation was monitored to ensure that a steadyrate was maintained throughout the experiment

In all lists there was a retention interval of 10 sec (the same re-tention interval as that used by Miles et al 1991) between presenta-tion of the final item and recall of the target items Participants weretold that they should try to remember the numbers by repeating themsilently to themselves during the retention interval A further tonesounded to mark the end of the retention interval This was the sig-nal for participants to commence recall of the list items They weretold to write the list items down in the order in which they had heardthem and that they were not allowed to change any earlier itemsonce they had written down a subsequent item Their recall sheetcontained six boxes for each list corresponding to the six serial po-

sitions They were allowed 12 sec to recall the digits before a furthertone sounded to tell them to stop recall There was then a gap of 5 secbefore the tone sounded again to signal the start of the next list

Irrelevant speech was presented through the left headphone spo-ken by a male voice Irrelevant speech comprised letters of the al-phabet being read out in a random order at the rate of approximatelyone and a half letters per second All 60 lists were associated with adifferent 6-sec segment of irrelevant speech For half of the partici-pants the irrelevant speech occurred during list presentation Forthese participants the irrelevant speech started exactly 1 sec afterpresentation of the first digit and continued until 1 sec after the finallist item had been presented The other half of the participants heardthe irrelevant speech during the last 6 sec of the retention intervalFor each list exactly the same token of irrelevant speech was usedregardless of whether it occurred during presentation or during theretention interval This ensured that the ldquodoserdquo (Bridges amp Jones1996) of the irrelevant speech was identical regardless of whether itwas presented at encoding or during the retention interval

ResultsA significance level of 05 (two-tailed) was established

for all statistical tests reported in this paper Overall meanperformance in Experiment 1 is summarized in Figure 2A three-way analysis of variance was performed on thenumber of items correctly recalled in serial order This re-vealed a significant main effect of irrelevant speech[F(128) = 395 MSe = 319] a significant main effect ofarticulatory suppression [F(138) = 2770 MSe = 561]but no effect of location of the irrelevant speech (F lt 1)The interaction between articulatory suppression and ir-relevant speech was significant [F(138) = 81 MSe =314] Neither of the remaining two-way interactions ap-proached significance The three-way interaction alsofailed to reach significance (F lt 1)

The significant interaction between articulatory sup-pression and irrelevant speech was further investigated bytests of simple main effects These revealed significant ef-

Figure 2 Mean number (plus standard error bars) of phonologically distinct andsimilar items correctly recalled per list (max 5 6) with articulatory suppression (AS)and without articulatory suppression (NoAS) in the presence and absence of irrele-vant speech (IS) in Experiment 1


fects of irrelevant speech both with articulatory suppres-sion [F(176) = 418 MSe = 316] and without articulatorysuppression [F(176) = 61 MSe = 316] The effects of ar-ticulatory suppression were significant both with irrele-vant speech [F(176) = 2259 MSe = 437] and without irrelevant speech [F(176) = 1353 MSe = 437] The inter-action between irrelevant speech and articulatory sup-pression seems to have occurred because the effects of ir-relevant speech are somewhat larger with rather than withoutarticulatory suppression (Figure 2) The most likely ex-planation is that the effects of irrelevant speech are beingreduced slightly in the no-suppression condition as a re-sult of a ceiling effect

DiscussionThe results of this experiment have replicated those of

the second and third experiments by Hanley and Broad-bent (1987) in showing significant effects of irrelevantspeech on auditorily presented items under conditions ofarticulatory suppression consistent with Salameacute and Bad-deleyrsquos (1982) phonological loop model Because audito-rily presented items enter the phonological store automat-ically they will be subject to interference from irrelevantspeech even under conditions in which participants per-form articulatory suppression The results are contrary tothe predictions of Neathrsquos (2000) feature model Neath ar-gued that articulatory suppression and irrelevant speechboth interfere with the representations of modality-independent features in primary memory traces Once ar-ticulatory suppression has affected these memory tracesirrelevant speech should cause them no further damage

The findings reported here extend those of Hanley andBroadbent (1987) by showing significant effects of irrel-evant speech under conditions of articulatory suppressioneven when the irrelevant speech was presented during theretention interval (Figure 2) Neath (2000) suggested thatthe effects of irrelevant speech under articulatory sup-pression that Hanley and Broadbent reported might havecome about because the irrelevant speech masked the per-ception of the auditorily presented target words If thissuggestion were true one would have expected that the ef-fects of irrelevant speech should have been greatly re-duced when the irrelevant speech was presented duringthe retention interval This is because there can be no per-ceptual masking when the irrelevant speech starts severalseconds after the target items have been presented Theabsence of any relationship between the location of the ir-relevant speech and the strength of the irrelevant speecheffect (Figure 2) shows that this is not the case If Neathrsquossuggestion were true there should also have been additiveeffects of irrelevant speech and articulatory suppressionwhen the irrelevant speech occurred at encoding but aninteraction between articulatory suppression and irrele-vant speech when the irrelevant speech occurred duringthe retention interval The complete absence of any three-way interaction shows that this is not the case The strengthof the irrelevant speech effect under articulatory suppres-sion is just as strong when the irrelevant speech occurs inthe retention interval as when it occurs during list presen-

tation (Figure 2) The results therefore further confirmone of the key predictions about the irrelevant speech ef-fect made by Salameacute and Baddeleyrsquos (1982) phonologicalloop model

It is well known that the presentation of a single item orldquosuffixrdquo immediately after the final item in a list can sig-nificantly reduce serial recall of auditorily presented lists(eg Crowder 1967) According to some (eg Crowder ampMorton 1969) the suffix interferes with the precategori-cal representation of the final list item within auditory sen-sory memory Is it conceivable therefore that the effect ofirrelevant speech in Experiment 1 came about because thespeech functioned as a suffix If so the feature modelcould accommodate the results from Experiment 1 This isbecause the feature model (see Surprenant et al 2000) as-sumes that a suffix interferes with the modality-dependentfeatures of the final list item (whereas irrelevant speechand articulatory suppression interfere with modality-independent features) However it seems unlikely that theeffects of irrelevant speech when presented during the re-tention interval acted simply as a suffix First Crowder(1978) reported that a suffix had little effect if presentedmore than 2 sec after the final list item but in Experiment 1there was a delay of 6 sec between the presentation of thefinal list item and the start of the irrelevant speech Secondthere is no evidence that the effects of a suffix can persistover a retention interval as long as that used in Experi-ment 1 (10 sec) For example Watkins and Todres (1980)showed that unless the participants were asked to performa secondary task during the retention interval a suffix ex-erted no effect on serial recall if there was a delay of 12 secbetween the end of the list and recall Finally MortonCrowder and Prussin (1971) showed that the effects of asuffix were substantially reduced if the targets were pre-sented to one ear and the suffix to the other Morton et alalso reported that a pure tone did not produce a suffix ef-fect Irrelevant speech effects on the other hand are pro-duced by sequences of changing-state auditory tones (Jonesamp Macken 1993) It would therefore be possible to inves-tigate this issue further by comparing the effects of irrel-evant tones and irrelevant words when presented duringthe retention interval following auditory presentation oflist items under conditions of articulatory suppression

EXPERIMENT 2

In Experiment 2 we investigated competing accountsof the combined effects of irrelevant speech and phonolog-ical similarity on serial recall As was pointed out earlier inthe paper some of the available evidence (Jones amp Macken1995 Larsen et al 2000 Salameacute amp Baddeley 1986) isconsistent with the phonological loop model because itdemonstrates a detrimental effect of phonological simi-larity on the recall of visually presented list items whenparticipants are exposed to irrelevant speech In otherstudies employing visual presentation of target materials(Colle amp Welsh 1976 Surprenant et al 1999) howeverthere was no effect of phonological similarity in the pres-ence of irrelevant speech These two studies are therefore


consistent with the feature model (Neath 2000) whichpredicts that under these circumstances the phonologicalsimilarity effect will be abolished by irrelevant speech

Salameacute and Baddeley (1986) offered an explanation forwhy phonological similarity effects sometimes do notoccur in the presence of irrelevant speech In their ownstudy they varied list length from 5 to 8 items per listThey found clear additive effects at List Lengths 5 and 6At List Lengths 7 and 8 the interaction between phono-logical similarity and irrelevant speech approached sig-nificance and at List Length 8 there was no main effect ofphonological similarity Salameacute and Baddeley (1986) sug-gested that at short list lengths their participants adoptedan articulatoryphonological rehearsal strategy that madeuse of the phonological loop Salameacute and Baddeley (1986)argued that when lists contain large numbers of phono-logically similar items however phonological coding islikely to be difficult and will be abandoned by many par-ticipants in favor of a visual or semantic strategy In sup-port of this account they reported that high-performingparticipants showed additive effects of phonological sim-ilarity and irrelevant speech even with a list length of eightitems They suggested that these individuals had greaterphonological loop capacity and were able to maintain aphonological coding strategy throughout the experiment

Colle and Welsh (1976) and Surprenant et al (1999)both used relatively long lists (eight items per sequence)Consequently Salameacute and Baddeleyrsquos (1986) account canexplain the absence of an effect of phonological similar-ity in the presence of irrelevant speech in these studiesalso by assuming that many participants abandoned aphonological coding strategy Salameacute and Baddeleyrsquos (1986)account is also consistent with the results of Larsen et al(2000) who showed additive effects of irrelevant speechand phonological similarity in two experiments that em-ployed relatively short sequences (six items per list)

Nevertheless there is no direct evidence to supportSalameacute and Baddeleyrsquos (1986) claim that significant ef-fects of phonological similarity are dependent on the useof a phonological coding strategy Nor is there direct evi-dence that a semantic strategy will abolish the effects ofphonological similarity In Experiment 2 therefore wesought to investigate these issues further by giving someof the participants explicit instructions on how theyshould encode the target items One group was told to usea phonological rehearsal strategy We assumed that thisgroup would use the phonological loop to rehearse the tar-get items On the basis of Salameacute and Baddeleyrsquos (1986)account it was therefore predicted that this group wouldshow additive effects of irrelevant speech and phonologi-cal similarity Another group was instructed to use a se-mantic strategy and it was assumed that they would notuse the phonological loop It was therefore predicted thatthis group would show no effect of irrelevant speech orphonological similarity A third group was given no instruc-tions (standard procedure in serial recall experiments) Weused list lengths of seven items per sequence because thisis the length at which Salameacute and Baddeleyrsquos (1986) re-sults suggested that some participants might abandon a

phonological strategy in the presence of irrelevant speechwhen items are phonologically similar It was predictedthat the performance of this group would resemble that ofthe participants in Colle and Welshrsquos (1976) and Sur-prenant et alrsquos (1999) studies in showing effects of phono-logical similarity only in quiet conditions

MethodParticipants Fifty-four students at Essex University took part in

the experiment and were tested individually They were either paidpound3 or received a course credit for taking part There were 16 partic-ipants in the phonological group 16 participants in the semanticgroup and 22 participants in the control group

Materials All lists of target items contained seven randomly or-dered items sampled without replacement from either a set ofphonologically similar letters (B C D G P T V) or a set of phono-logically different letters (H J R S L Y Z) The irrelevant speechconsisted of a tape recording of a passage from Bleak House byCharles Dickens read by a native English speaker

Design and Procedure Irrelevant speech (present vs absent)and phonological similarity (similar vs different) were both manip-ulated within subjects Each participant received a total of 48 lists di-vided into two blocks of 24 lists each Participants were exposed toirrelevant speech during one block of lists and one block was pre-sented in silence Half of the participants heard irrelevant speechduring the first block of lists and half of the participants heard ir-relevant speech during the second block of lists Irrelevant speechwas presented through a loudspeaker during list presentation andduring recall of list items Participants were told to ignore the irrel-evant speech The first four lists during each block were deemedpractice lists Half of the critical and half of the practice lists in eachblock contained phonologically similar letters and half containedphonologically different letters The order of the phonologically sim-ilar and different lists within each block was randomized

Target items were presented one at a time in uppercase on thescreen of an Apple Macintosh computer At the start of each list theword ready appeared for 075 sec on the computer screen Thescreen then went blank for 05 sec before the first target letter ap-peared in 48-point Geneva font Each letter remained on the screenfor 10 sec with an interstimulus interval of 05 sec After the finalletter disappeared from the screen a tone sounded for 025 sec Par-ticipants were then allowed 12 sec to attempt to recall the list itemsWhen the 12 sec was over the tone sounded and participants wereasked to look at the screen once again for the start of the next listRecall instructions were the same as those used in Experiment 1

Before the experiment began the participants were given one ofthree different sets of instructions The participants in the semanticgroup were told to try and remember the letters by turning each oneinto a word that started with that letter They were then told to at-tempt to create a meaningful sentence from the words that they hadgenerated maintaining the order in which the letters had been pre-sented They could then use the sentence to help them recall the orderof the letters at test The phonological group was told to rememberthe letters by repeating them subvocally until they had to be recalledThe control group was given no instructions as to how to rememberthe items At the end of the experiment all participants were askedto describe the strategies that they had used to remember the listitems All of the participants in the phonological group indicatedthat they had used a phonological strategy throughout the experi-ment Sixty-nine percent of the participants in the semantic group re-ported consistent use of the strategy they had been asked to employThe strategies employed in the control group are discussed below

ResultsOverall mean performance in Experiment 2 is summa-

rized in Figure 3


An analysis of performance in the phonological grouprevealed a significant main effect of irrelevant speech[F(115) = 134 MSe = 4358] and a significant main ef-fect of phonological similarity [F(115)= 4835 MSe = 504]The interaction between irrelevant speech and phonolog-ical similarity was not significant [F(115) = 13 MSe =376] Planned comparisons (tests of simple main effects)revealed significant effects of phonological similarity inquiet conditions [F(115)= 472 MSe = 335] and in the pres-ence of irrelevant speech [F(115) = 166 MSe = 545]

An analysis of performance in the semantic group re-vealed no significant main effect of irrelevant speech[F(115) = 42 MSe = 374 p = 06] or of phonologicalsimilarity (F lt 1) The interaction between irrelevantspeech and phonological similarity was not significant(F lt 1) Planned comparisons (tests of simple main effects)revealed no significant effect of phonological similarityin quiet conditions or in the presence of irrelevant speech(both Fs lt 1)

An analysis of performance in the control group re-vealed a significant main effect of irrelevant speech[F(121) = 43 MSe = 621] and a significant main effectof phonological similarity [F(121) = 44 MSe = 1143]There was also a significant interaction between irrelevantspeech and phonological similarity [F(121) = 48 MSe =220] Planned comparisons (tests of simple main effects)revealed significant effects of phonological similarity inquiet conditions [F(121) = 64 MSe = 827] but no sig-nificant effect of phonological similarity in the presenceof irrelevant speech [F(121) = 14 MSe = 536]

DiscussionThe results of this experiment revealed strikingly dif-

ferent patterns of performance in the three groups of par-ticipants When no encoding instructions were given therewas a significant effect of phonological similarity in quietconditions but not in the presence of irrelevant speechWhen participants were given phonological coding in-structions there were significant effects of phonologicalsimilarity even in the presence of irrelevant speech Withsemantic instructions there was no effect of phonologicalsimilarity even in quiet conditions

The fact that there were additive effects of phonologi-cal similarity and irrelevant speech when participantswere told to use articulatoryphonological rehearsal pro-vides strong support for the phonological loop model(Salameacute amp Baddeley 1982) According to this model vi-sually presented material will enter the phonological storewhen it is articulated even in the presence of irrelevantspeech Once they have entered the phonological storephonologically similar items will be more diff icult tomaintain than phonologically different items Both typesof item will also suffer interference from the irrelevantspeech Consequently irrelevant speech and phonologicalsimilarity should make separate contributions to reducingserial recall performance Conversely the feature model(Neath 2000) predicts that irrelevant speech should abol-ish the phonological similarity effect with visual presen-tation of targets The fact that there was a phonologicalsimilarity effect in the presence of irrelevant speech istherefore inconsistent with the feature model

Figure 3 Mean number ( 6 SE ) of phonologically different and phonologically similar items cor-rectly recalled per list (max 5 7) by the three experimental groups in quiet conditions and in thepresence of irrelevant speech in Experiment 2


The semantic group showed no effect of phonologicalsimilarity and a small nonsignificant effect of irrelevantspeech with the majority of participants reporting thatthey had successfully used the semantic strategy that theyhad been asked to employ Such an outcome shows that wheninstructed many individuals can readily adopt alternativestrategies that counteract the effects of variables that havelong been known to interfere with serial recall performanceThe performance of the semantic group in Experiment 2therefore confirms the claim (Baddeley 2000 Salameacute ampBaddeley 1986) that the effects of phonological similar-ity will disappear if participants switch to a semantic strat-egy as list length increases

The finding that irrelevant speech removed the phono-logical similarity effect when participants were given noencoding instructions is similar to the results of Colle andWelsh (1976) and Surprenant et al (1999) Because thisresult is quite different from that observed in the phono-logical group it must be concluded that at least some ofthe participants in the control group were not using articulatoryphonological rehearsal strategies Furthersupport for this conclusion emerged when the participantsin the control group were asked to outline the strategiesthey had used to remember the items Whereas all of theparticipants in the phonological group reported using aphonological strategy a wide range of different strategieswas described by the control group Six participants (27)in the control group reported that they consistently usedan articulatoryphonological strategy An analysis of theperformance of these 6 participants revealed a significanteffect of phonological similarity both in the presence[F(15) = 107 MSe = 629] and absence [F(15) = 71 MSe =629] of irrelevant speech This suggests that these indi-viduals were indeed using phonological coding Four par-ticipants (18) reported consistent use of a semantic strat-egy that involved converting the letters into words andmaking sentences or phrases from the words These 4 par-ticipants showed no evidence whatsoever of a phonologi-cal similarity effect performing slightly though non-significantly better overall on the phonologically similaritems Fourteen percent reported using a mixture of phono-logical and semantic strategies on different lists with se-mantic strategies being used more frequently on the phono-logically similar items Fourteen percent reported changingthe seven letters into seven words that started with the cor-responding target letter but instead of creating a mean-ingful sentence they then repeated these words sub-vocally Twenty-seven percent reported changing theletters into a smaller number of words or nonsense wordsby adding vowels between the target letters For examplethe sequence Y R J S L H Z might be converted to twononsense words such as YARJ SALHAZ These two wordswould then be repeated subvocally The sequence B C DG P V T might be converted to BOC DIG PRIVATE whichwould then be repeated Because they realized that the tar-get letters were all consonants these participants reportedthat they were able to avoid writing down vowels at recall

Clearly such a strategy would remove the phonologicalsimilarity effect because the participant would be unlikely

to generate phonologically similar words The use of thisstrategy and the semantic strategy fits well with what waspredicted by Salameacute and Baddeley (1986) when they wereattempting to explain why the effects of irrelevant speechand phonological similarity tended to be additive at shortlist lengths and to interact at longer list lengths Experiment 2provides direct evidence that when lists contain sevenitems per sequence some participants will adopt a strat-egy that negates the effects of variables such as phono-logical similarity and irrelevant speech These results arealso consistent with Baddeleyrsquos (1966) finding that withlong sequences serial recall is affected by semantic ratherthan phonological similarity between list items

The relatively good performance of the semantic groupmight bring into question whether or not a phonologicalstrategy is an effective means of retaining verbal sequencesin short-term memory It must be borne in mind howeverthat the experimental manipulations that were employedin this experiment (phonologically similar lists exposureto irrelevant speech the use of supra-span lists) are specif-ically designed to put the phonological loop under extremestress Nevertheless it is surprising that the semanticgroup performed so much better than the phonologicalgroup and the effectiveness of semantic strategies in ser-ial recall is well worth further investigation in the future

GENERAL DISCUSSION

The results of both Experiments 1 and 2 have producedresults that conflict with the predictions derived fromNeathrsquos (2000) feature model Experiment 1 revealed ad-ditive effects of irrelevant speech and articulatory sup-pression with auditory presentation of list items Neathclaimed that the only way that such a result could occurwould be if the irrelevant speech masked the presentationof the target items In Experiment 1 however similar re-sults were obtained even when the irrelevant speech waspresented after the target items In Experiment 2 therewere additive effects of irrelevant speech and phonologi-cal similarity with visual presentation as long as partici-pants were asked to use phonological coding strategiesThe feature model however predicts that irrelevantspeech should abolish the phonological similarity effectwith visual presentation when participants are usingphonological encoding3 Overall therefore these resultscast doubt on Neathrsquos (p 419) claim that the feature modelldquocorrectly predicts the interactions between irrelevantspeech and other factors long considered important in thestudy of immediate memoryrdquo

The basic problem is Neathrsquos (2000) claim that irrelevantspeech and articulatory suppression have broadly similareffects on serial recall Hence it incorrectly predicts nocombined effect of irrelevant speech and articulatory sup-pression in Experiment 1 and incorrectly predicts an in-teraction between phonological similarity and irrelevantspeech in the phonological group in Experiment 2 The re-sults however show that irrelevant speech can behavequite differently from articulatory suppression For exam-ple in the phonological group in Experiment 2 there were


significant effects of phonological similarity on visuallypresented sequences in the presence of irrelevant speechIn the presence of articulatory suppression however it iswell known that the effect of phonological similarity onvisually presented sequences is abolished (eg Baddeleyet al 1984 Longoni et al 1993 Murray 1967 1968)

Conversely additive effects of irrelevant speech and ar-ticulatory suppression with auditory presentation of tar-get items (Experiment 1) and additive effects of irrelevantspeech and phonological similarity (Experiment 2) withvisual presentation of target items follow exactly the pre-dictions of the phonological loop model (eg Baddeley2000) The phonological loop model predicts effects of ir-relevant speech in the presence of articulatory suppressionas long as target materials are presented auditorily (Ex-periment 1) This is because articulatory suppression willnot prevent auditorily presented material from enteringthe phonological loop where it will become subject to in-terference from irrelevant speech There can be phono-logical similarity effects on the serial recall of visuallypresented items even in the presence of irrelevant speech(Experiment 2) because irrelevant speech affects the stor-age of verbal materials within the phonological store butdoes not prevent access of visually presented materials tothe phonological store according to the phonological loopmodel The demonstration that phonological similarity ef-fects are removed with semantic encoding instructions isalso consistent with claims made by Salameacute and Badde-ley (1986) and Baddeley (2000) This finding providessupport for Baddeleyrsquos (2000) suggestion that some pre-vious studies (Colle amp Welsh 1976 Surprenant et al1999) failed to obtain phonological similarity effects inthe presence of irrelevant speech because some partici-pants abandoned phonological encoding strategies as listlength increased Consequently it appears that the phono-logical loop model provides a superior account of the ir-relevant speech effect than does the feature model4

The phonological loop model can also offer a possibleexplanation of why there are additive effects of irrelevantspeech and phonological similarity with auditory presen-tation at list lengths in which nonsignificant effects ofphonological similarity are observed with visual presen-tation (Surprenant et al 1999) Because phonologicalpresentation automatically registers target items in thephonological store it seems reasonable to conclude thatrelative to visual presentation it will increase the proba-bility that participants will employ phonological codingSuch an account could be tested by repeating Surprenantet alrsquos (1999) study and asking participants to outline thestrategies that they used We predict that phonologicalstrategies would be reported significantly more frequentlywith auditory presentation than with visual presentationat the sequence length that Surprenant et al (1999) used

One of the most attractive aspects of Neathrsquos (2000)feature model is the fact that it can be implemented as acomputational model The model therefore contains an ex-plicit account of the way in which serial order is retainedPerforming simulations makes it possible to make precisepredictions as to how the feature model will behave in the

presence of one or more of the variables that are known toinfluence serial recall performance It is also possible topredict the shape that the serial position curve will takeComputational models of the phonological loop havebeen developed in recent years (Burgess amp Hitch 1999Henson 1998 Page amp Norris 1998) but none of them haveas yet attempted to explain the irrelevant speech effectOur results suggest that the phonological loop model pro-vides an appropriate framework for understanding a num-ber of aspects of the irrelevant speech effect but it is clearthat a lot of important theoretical work remains to be done

REFERENCES

Baddeley A D (1966) Short-term memory for word sequences as afunction of acoustic semantic and formal similarity Quarterly Jour-nal of Experimental Psychology 18 362-365

Baddeley A D (1968) How does acoustic similarity influence short-term memory Quarterly Journal of Experimental Psychology 20249-264

Baddeley A D (1986) Working memory Oxford Oxford UniversityPress Clarendon Press

Baddeley A D (1990) Human memory Theory and practice Lon-don Erlbaum

Baddeley A D (1992) Is working memory working Quarterly Jour-nal of Experimental Psychology 44A 1-31

Baddeley A D (2000) The phonological loop and the irrelevantspeech effect Some comments on Neath (2000) Psychonomic Bul-letin amp Review 7 544-549

Baddeley A D Lewis V amp Vallar G (1984) Exploring the ar-ticulatory loop Quarterly Journal of Experimental Psychology 36A233-252

Bridges A M amp Jones D M (1996) Word dose in the disruption ofserial recall by irrelevant speech Phonological confusion or changingstate Quarterly Journal of Experimental Psychology 49A 919-939

Burgess N amp Hitch G J (1999) Memory for serial order A networkmodel of the articulatory loop Journal of Memory amp Language 31429-460

Colle H A amp Welsh A (1976) Acoustic making in primary mem-ory Journal of Verbal Learning amp Verbal Behavior 15 17-32

Conrad R amp Hull A J (1964) Information acoustic confusion andmemory span British Journal of Psychology 55 429-432

Crowder R G (1967) Prefix effects in immediate memory CanadianJournal of Psychology 21 450-461

Crowder R G (1978) Mechanisms of backward masking in the stim-ulus suffix effect Psychological Review 85 502-504

Crowder R G amp Morton J (1969) Pre-categorical acoustic storage(PAS) Perception amp Psychophysics 5 365-373

Ellermeier W amp Zimmer K (1997) Individual differences in sus-ceptibility to the irrelevant speech effect Journal of the AcousticalSociety of America 102 2191-2199

Gathercole S E amp Baddeley A D (1993) Working memory andlanguage Hove UK Erlbaum

Hanley J R (1997) Does articulatory suppression remove the irrele-vant speech effect Memory 5 423-431

Hanley J R amp Broadbent C (1987) The effects of unattendedspeech on serial recall following auditory presentation British Jour-nal of Psychology 78 287-297

Henson R N A (1998) Short-term memory for serial order The start-end model Cognitive Psychology 36 73-117

Jones D M (1994) Disruption of memory for lip read lists by irrele-vant speech Further support for the changing state hypothesis Quar-terly Journal of Experimental Psychology 47A 143-160

Jones D M amp Macken W J (1993) Irrelevant tones produce an ir-relevant speech effect Implications for phonological coding in work-ing memory Journal of Experimental Psychology Learning Mem-ory amp Cognition 19 369-381

Jones D M amp Macken W J (1995) Phonological similarity in theirrelevant speech effect Within or between stream similarity Journal


of Experimental Psychology Learning Memory amp Cognition 21103-115

Jones D M Madden C A amp Miles C (1992) Privileged access byirrelevant speech to short-term memory Quarterly Journal of Exper-imental Psychology 44A 645-659

Larsen J D Baddeley A D amp Andrade J D (2000) Phonolog-ical similarity and irrelevant speech Implications for models of short-term verbal memory Memory 8 145-158

Longoni A M Richardson J T E amp Aiello A (1993) Articula-tory rehearsal and phonological storage in working memory Memoryamp Cognition 21 11-22

Miles C Jones D M amp Madden C A (1991) Locus of the irrele-vant speech effect in short-term memory Journal of ExperimentalPsychology Learning Memory amp Cognition 17 578-584

Morton J Crowder R G amp Prussin H (1971) Experiments withthe stimulus suffix effect Journal of Experimental Psychology 91169-190

Murray D J (1967) The role of speech responses in short-term mem-ory Canadian Journal of Psychology 21 263-276

Murray D J (1968) Articulation and acoustic confusability Journalof Experimental Psychology 22 679-684

Nairne J S (1990) A feature model of immediate memory Memoryamp Cognition 18 251-269

Neath I (2000) Modeling the effects of irrelevant speech on memoryPsychonomic Bulletin amp Review 7 403-423

Neath I Farley L A amp Surprenant A M (in press) Directly as-sessing the relationship between irrelevant speech and articulatorysuppression Quarterly Journal of Experimental Psychology

Neath I Surprenant A M amp LeCompte D C (1998) Irrelevantspeech eliminates the word length effect Memory amp Cognition 26343-354

Page M P A amp Norris D (1998) The primacy model A new modelof immediate serial recall Psychological Review 105 761-781

Salameacute P amp Baddeley A D (1982) Disruption of short-term mem-ory by irrelevant speech Implications for the structure of workingmemory Journal of Verbal Learning amp Verbal Behavior 21 150-164

Salameacute P amp Baddeley A [D] (1986) Phonological factors in STMSimilarity and the unattended speech effect Bulletin of the Psycho-nomic Society 24 263-265

Surprenant A M LeCompte D C amp Neath I (2000) Manipula-tions of irrelevant information Suffix effects with articulatory sup-

pression and irrelevant speech Quarterly Journal of ExperimentalPsychology 53A 325-348

Surprenant A M Neath I amp Lecompte D C (1999) Irrelevantspeech phonological similarity and presentation modality Memory7 405-420

Watkins M J amp Todres A K (1980) Suffix effects manifest andconcealed Further evidence for a 20-second echo Journal of VerbalLearning amp Verbal Behavior 19 46-53

NOTES

1 The findings of Jones and Macken (1995) have sometimes been in-terpreted (eg Larsen et al 2000) as being similar to those of Colle andWelsh (1976) In fact Jones and Macken (Experiment 4) reported sig-nificant main effects of phonological similarity and irrelevant speech to-gether with a nonsignificant interaction between these variables

2 Because articulatory suppression reduces recall more substantiallythan irrelevant speech the value of a separate parameter labeled ldquoatten-tionrdquo is also reduced in Neathrsquos (2000) simulations of the effects of ar-ticulatory suppression

3 The version of the feature model that was implemented by Neath(2000 see p 420) assumed that ldquothe role of phonological information isconstantrdquo However Nairnersquos (1990) original model included a parame-ter designed to weight particular feature comparisons (see Neath Farleyamp Surprenant in press) If one were to use this parameter to weightphonological features more heavily under phonological than under se-mantic coding instructions then it might be possible that the featuremodel could simulate the performance of the phonological group in Fig-ure 3

4 Nevertheless there are some remaining weaknesses associated withthe phonological loop modelrsquos explanation of the irrelevant speech ef-fect There is no account of the way in which serial order is representedwithin the model and the fact that there is no additional effect of irrele-vant speech when the speech and the targets are phonemically similarrequires explanation See Baddeley (2000) for a recent discussion ofthese issues

(Manuscript received November 30 2001 accepted for publication February 25 2002)


reported by Hanley (1997) Salameacute and Baddeley (1982)claimed the reason for this interaction is that concurrentarticulation isolates visually presented material from thephonological store Although performance will be loweras a consequence it will not be further reduced by expo-sure to irrelevant speech

It follows from the phonological loop model that if thetarget material is presented auditorily then irrelevantspeech should reduce serial recall performance under ar-ticulatory suppression This is because the auditorily pre-sented list items will enter the phonological store auto-matically As a consequence list items will be subject tointerference from irrelevant speech even though concurrentarticulation is taking place This prediction was supportedby Hanley and Broadbent (1987) They showed additiveeffects of irrelevant speech and articulatory suppressionon the recall of auditorily presented lists of items in twoout of three experiments even when articulatory suppres-sion continued during presentation and recall We investi-gate this issue further in Experiment 1

The phonological loop model also provides an accountof the relationship between the effects of irrelevant speechand the effects of phonological similarity (Baddeley1966 1968 Conrad amp Hull 1964) on serial recall It isclaimed (eg Baddeley Lewis amp Vallar 1984) that therepresentations of phonologically similar items cause in-terference to one another within the phonological storeConsistent with this claim a number of studies haveshown that with visual presentation of list items the ef-fects of phonological similarity are also abolished by ar-ticulatory suppression (eg Baddeley et al 1984 Lon-goni Richardson amp Aiello 1993 Murray 1967 1968)Because both the irrelevant speech effect and the phonolog-ical similarity effect reflect the operation of the phono-logical store it follows from the phonological loop model

that the effects of phonological similarity and irrelevantspeech should be additive In other words phonologicallysimilar items should be recalled worse than phonologi-cally different items even in the presence of irrelevantspeech Although this prediction has been supported insome studies (Jones amp Macken 19951 Larsen Baddeleyamp Andrade 2000 Salameacute amp Baddeley 1986) the evi-dence is mixed Neither Colle and Welsh (1976) nor Sur-prenant Neath and LeCompte (1999) observed a phono-logical similarity effect in the presence of irrelevant speechWe investigate this issue further in Experiment 2

Feature Model Accountof the Irrelevant Speech Effect

Neathrsquos (2000) alternative model of the irrelevantspeech effect is based on Nairnersquos (1990) feature modelThe feature model assumes that each item from a list isrepresented in memory as a discrete set of features that areset to either 1 or 21 Every item is represented in both pri-mary memory and secondary memory Although sec-ondary memory traces are considered to be intact tracesin primary memory are subject to interference from lateritems The presentation of a new item that shares a featurevalue with one or more items from earlier in the list willhave the effect of altering the value of that feature from 1or ndash1 to zero in the memory traces of all the previouslypresented items This renders the feature useless for thepurposes of recalling the earlier items At recall eachtrace in primary memory is used to access the most simi-lar trace in secondary memory with the trace that hasgreatest feature overlap with the trace in primary memorybecoming selected for recall As features become lostfrom the trace of the item in primary memory so the prob-ability increases that an incorrect item will be recalled

The features that represent list items are either modality-independent or modality-dependent features The formerare activated regardless of modality of presentationwhereas different modality-dependent features are acti-vated when an item is presented visually from those thatare activated when it is presented auditorily It is assumedthat there are more modality-dependent features associ-ated with the auditory modality than with the visual modal-ity (hence the modality effect in serial recall)

The Effects of Irrelevant Speech andArticulatory Suppression

According to Neath (2000) irrelevant speech and artic-ulatory suppression reduce recall in broadly similar waysThe word that is repeated by the participant during articu-latory suppression causes additional interference to previ-ously presented list items Some of the modality-independentfeatures in primary memory are replaced by features ofthe word that is articulated a process referred to as featureadoption To simulate this effect half of the modality-independent features in each trace are set to a constantvalue (11) Feature adoption is also the cause of the in-terference that is brought about by irrelevant speechSome of the features of the words in the irrelevant speechstream replace features of the traces of the target items in

Figure 1 A representation of the phonological loop modeltaken from Gathercole and Baddeley (1993) Speech inputs (egirrelevant speech and auditorily presented list items) gain auto-matic entry to the phonological store Nonspeech inputs (eg vi-sually presented list items) require articulatory rehearsal inorder to access the phonological store and are therefore cut offfrom the phonological store by articulatory suppression Irrele-vant speech will interfere with the representations of list items ifand only if they are being held within the phonological store Theform in which visually presented words or letters are stored undersuppression is not specified by the model







EXPERIMENT 1
























EXPERIMENT 2















rized in Figure 3















GENERAL DISCUSSION









REFERENCES











































NOTES












EXPERIMENT 1
























EXPERIMENT 2















rized in Figure 3















GENERAL DISCUSSION









REFERENCES











































NOTES
























EXPERIMENT 2















rized in Figure 3















GENERAL DISCUSSION









REFERENCES











































NOTES



















rized in Figure 3















GENERAL DISCUSSION









REFERENCES











































NOTES




















GENERAL DISCUSSION









REFERENCES











































NOTES












REFERENCES











































NOTES


























NOTES






Irrelevant speech, articulatory suppression, and ... · PDF fileA test of the phonological...

Documents

Transcript of Irrelevant speech, articulatory suppression, and ... · PDF fileA test of the phonological...