Generalization of tactile perceptual skills to new contextfollowing tactile-alone word recognition training with theTickle Talker™

Karyn L. GalvinCRC for Cochlear Implant and Hearing Aid Innovation, 384-388 Albert Street, East Melbourne 3002,Australia and The Bionic Ear Institute, 384-388 Albert Street, East Melbourne 3002, Australia

Peter J. BlameyDepartment of Otolaryngology, The University of Melbourne, 2nd Floor, RVEEH, 32 Gisborne Street,East Melbourne 3002, Australia

Robert S. C. CowanCRC for Cochlear Implant and Hearing Aid Innovation, 384-388 Albert Street, East Melbourne 3002,Australia; The Bionic Ear Institute, 384-388 Albert Street, East Melbourne 3002, Australia;and Department of Otolaryngology, The University of Melbourne, 2nd Floor, RVEEH, 32 Gisborne Street,East Melbourne 3002, Australia

Michael OerlemansDepartment of Otolaryngology, The University of Melbourne, 2nd Floor, RVEEH, 32 Gisborne Street,East Melbourne 3002, Australia

Graeme M. ClarkCRC for Cochlear Implant and Hearing Aid Innovation, 384-388 Albert Street, East Melbourne 3002,Australia; The Bionic Ear Institute, 384-388 Albert Street, East Melbourne 3002, Australia;and Department of Otolaryngology, The University of Melbourne, 2nd Floor, RVEEH, 32 Gisborne Street,East Melbourne 3002, Australia

~Received 11 May 1999; revised 26 May 2000; accepted 28 August 2000!

The Tickle Talker™ is an electrotactile speech perception device. Subjects were evaluated using thedevice in various tactile-alone and tactile–visual contexts to assess the generalization to othercontexts of tactile-alone perceptual skills. The subjects were from a group of six normally hearingsubjects who had previously received 12 to 33 h of tactile-alone word recognition training and hadlearned an average vocabulary of 50 words@Galvin et al., J. Acoust. Soc. Am.106, 1084–1089~1999!#. The tactile-alone evaluation contexts were sentences, unfamiliar talkers, and untrainedwords. The tactile–visual evaluation contexts were closed-set words, open-set words, and open-setsentences. Tactile-alone perceptual skills were generalized to unfamiliar speakers, sentences, anduntrained words, though scores indicated that generalization was not complete. In contrast, thegeneralization of skills to tactile–visual contexts was minimal or absent. The potential value oftactile-alone training for hearing-impaired users of the Tickle Talker™ is discussed. ©2000Acoustical Society of America.@S0001-4966~00!00612-3#

PACS numbers: 43.66.Wv, 43.71.Es, 43.71.Ky, 43.71.Ma@DOS#

enfonaire

dicoroe

on

icl

per

evelInco-ess93;

ice,i-p-

ice,p-

nts,d fort aho

I. INTRODUCTION

Tactile devices supply speech information to supplemthat available via audition and vision, and are one optionimproving the sound awareness and speech perceptiopeople with a severe-profound or profound hearing impment. Multichannel tactile devices generally provide timintensity, and spectral information that can be used tocriminate speech features, and to recognize vowels andsonants in closed sets. Studies in the literature have repovariation in the level of enhancement demonstratedspeechtracking evaluations, with relatively small improvments of 5 to 13 words per minute~wpm! being commonwhen the tactile was combined with auditory, visual,auditory-plus-visual information~see, for example, Cowaet al., 1988, 1989, 1990; Kishon-Rabinet al., 1996; Plant,1992, 1998; Reed and Delhorne, 1995; and summary artby Bernstein, 1992; Weisenberger, 1995!. Superior

2969 J. Acoust. Soc. Am. 108 (6), December 2000 0001-4966/2000/1

trof-,s-n-

tedn-

r

es

speechtracking enhancement rates of over 40 wordsminute have also been reported~Weisenbergeret al., 1989!.Despite the potential demonstrated by these results, the lof clinical usage of tactile devices remains relatively low.contrast, an alternative speech perception device, thechlear implant, has achieved widespread clinical succ~see, for example, Geers and Moog, 1995; Kessler, 19Osbergeret al., 1996; Parkinsonet al., 1998; Stalleret al.,1995!. An implant has clear advantages over a tactile devincluding utilizing the natural auditory neural pathway, typcally providing an individual with a benefit to speech percetion greater than that to be expected from a tactile devand providing a realistic possibility for device-alone percetion of connected speech.

Given the advantages and success of cochlear implasome clinicians and researchers have questioned the neetactile devices. However, it is important to recognize thacochlear implant is not the answer for every person w

296908(6)/2969/11/$17.00 © 2000 Acoustical Society of America

bjects

2970 J. Acoust. S

TABLE I. Details of devices used in previous studies examining the post-training performance of sutrained in tactile-alone word recognition.

Reference Lynchet al.,1989

Lynch et al.,1988, 1989

Engelmann andRosov, 1975

Brookset al.,1986a, 1986b

Device Tactaid II Tacticon 1600 Oregon vocoder Queen’s Universityvocoder

Stimulation vibrotactile electrotactile vibrotactile vibrotactileSite wrist abdomen arm armChannels 2 16 23 16

m

unlen-

paiorgina

nticeebe

alysnveraeb

prd’sil

usi

tchlyh

lyextioer

tlo

ouinp

anm

klel-o-o-

ofrs

toor-f 50rd

keton-in-

es-e

orber-the

in ard

ofon

the, itichoneex-ne

on-ed

theg-x-ch

meets the audiological criteria for candidature. For sothere are medical contraindications to surgery, such asobliterated cochlea, a recurring middle-ear infection, orsuitability for general anesthetic. For others with multiphandicaps, their potential to cooperate with device fitting ato utilize the information provided may be difficult to predict. Even if candidacy is established, some adults andents of candidates do not wish to proceed with the operatReasons for this decision include unwillingness to undesurgery, the likely destruction of any remaining hearingthe implanted ear, the fact that performance predictionsbased on general group experience and cannot be guarafor an individual, and the cost of the surgery and the devParents may also agree with the view of some in the dcommunity that hearing impairment is not a disability to‘‘corrected’’ ~Lane and Bahan, 1998; Tucker, 1998!.

For individuals who cannot or do not wish to obtaincochlear implant, a tactile device is an alternative for supping supplementary acoustic information. Tactile device umay also have other important outcomes. Positive experiewith a tactile device motivates some individuals to improtheir communication, perhaps through participating in aurehabilitation classes, increased social participation, or invtigation of cochlear implant candidacy. Tactile device useestablished pediatric cochlear implant candidates mayvide useful information relating to such factors as the chilability to cooperate with programming and to learn to unew speech information, and the commitment of the chand family to session attendance and consistent device

Given that, for the foreseeable future, tactile devices wbe used by some hearing-impaired people, it is importanestablish what type of training is most likely to help suusers to learn to obtain and utilize tactile information quickand effectively. The importance of training has been empsized previously~see, for example, Alca´ntara, 1991; Galvinet al., 1993; Plant, 1995; Oller, 1995!; however, the optimumconfiguration of training remains unclear. Unfortunateclarifying the effective components of training is a compltask as many interacting factors influence speech percepThe limited number of hearing-impaired tactile device uswho are potential subjects for such research also makestask more difficult. One compromise approach is to emphearing subjects in studies examining limited aspectstraining. With continued research each such study shocontribute to an overall picture of the most appropriate traing for hearing-impaired tactile device users. One such scific area of investigation has been the value of unimodalcombined-modality training, and this topic is further exained in the present study.

oc. Am., Vol. 108, No. 6, December 2000

ean-

d

r-n.o

reeed.

af

-ece

ls-yo-sede.

llto

a-

,

n.sheyf

ld-e-d

-

The subjects in the present study used the TicTalker™ electrotactile device, which was originally deveoped by The University of Melbourne’s Department of Otlaryngology for use in the preoperative evaluation of cchlear implant candidates~Blamey and Clark, 1985!. Earliertraining studies with the device considered some effectsproviding unimodal and bimodal training to device use~Alcantaraet al., 1993; Oerlemans and Blamey, 1998!. In arecent investigation of the potential for naive subjectslearn a vocabulary using the Tickle Talker™ alone, six nmally hearing subjects learned to recognize an average owords when provided with 12 to 33 h of tactile-alone worecognition training~Galvin et al., 1999!. As discussed pre-viously ~Galvin et al., 1993!, some basic tactile-alone woris included in the Tickle Talker™ training program, with thaim of introducing the tactile signal, allowing the userfocus on the tactile information only, and developing an uderstanding that the tactile signal provides useful speechformation. However, the results of Galvinet al.’s ~1999!tactile-alone word recognition training study raise the qution of whether providing further tactile-alone training at thword level would be a productive use of training time fhearing-impaired device users. Such training would onlyworthwhile if the skills learned during training were genealized to other contexts which were representative ofvariation encountered in daily communication.

Previous studies have evaluated the performancevariety of contexts of subjects trained in tactile-alone worecognition using other tactile devices~the details of thesedevices are presented in Table I and further discussiontactile devices in general is provided in the Method secti!~Engelmann and Rosov, 1975; Reedet al., 1982; Brookset al., 1986a, 1986b; Lynchet al., 1988, 1989!. As the mainaim of these studies varied, the subjects in some studies~En-gelmann and Rosov, 1975; Reedet al., 1982; Lynchet al.,1989! gained experience or received training other thantactile-alone word recognition training. As a consequencewas not always possible to determine the extent to whperformance in other contexts was due to the tactile-altraining. The present study involved subjects whose onlyperience of tactile speech information was the tactile-aloword recognition training reported in Galvinet al. ~1999!.Given this, the performance of these subjects in other ctexts can be attributed to the generalization of skills learnduring the tactile-alone word recognition training.

The general aim of the present study was to considervalue of tactile-alone word recognition training for hearinimpaired users of the Tickle Talker™ by evaluating the fleibility of the tactile perceptual skills developed through su

2970Galvin et al.: Generalization of tactile perceptual skills

tungteileinth-s

nle

nlpreiyp

ripeedthore

heni-t 5utiu

l o

re

gthatilert

k20

orosthut

rg

tion-ant

iond-nsity

an therveionn ofchthe

anthe

atheodele-da-of

p to

lus

rallgh-dec-d asla-The300,

y.in-

em-y, ofin-re-ne

dsing

d nole-in

the

training. The specific aim was to assess whether percepskills acquired during tactile-alone word recognition trainiusing the Tickle Talker™ were generalized to other selectactile-alone and tactile–visual contexts. The other tactalone contexts were the perception of sentences, untrawords, and words presented by unfamiliar talkers. The otactile–visual contexts were the perception of closedwords and open-set words and sentences.

II. METHOD

A. Design

Four adults with normal hearing completed evaluatioof tactile-alone and tactile–visual perception with the TickTalker™ using a variety of test materials. The subjects’ oexperience with tactile speech perception was an earliergram of tactile-alone word recognition training, so that thperformance in other contexts could be attributed to this tof training.

B. Tactile devices and the Tickle Talker™

As background information to the subsequent desction of the Tickle Talker™, and to allow comparison of thTickle Talker™ with other tactile devices referred to in thDiscussion section, some basic points relating to tactilevices in general will be reviewed here. Characteristics oftactile sense are important in determining the type of infmation to be presented via a tactile device. Verrillo and Gscheider~1992! reviewed research into perception via tsense of touch. These authors concluded that the ski‘‘poor’’ at discriminating frequency differences, that the vbrotactile system has a comfortable intensity range of jusdB, and that masking of simultaneous or consecutive stimwas a significant issue for tactile device developers. A tacdevice usually consists of a microphone to collect the acotic input, a speech processor to process~if necessary! andrelay the signal, and transducer~s! to present the informationto the skin. Devices are commonly classified by the leveprocessing applied to the acoustic input~bandpass filter orspeech feature selection!, the site of stimulation, the numbeof transducers used to present information, and the typstimulation ~electrotactile or vibrotactile!. In devices em-ploying a speech-feature selection approach to processinacoustic input, the system extracts particular features ofacoustic signal for presentation to the user. In the bandpfilter or vocoder approach, the input passes through a sefilters, with each filter corresponding to a particular tacttransducer. This latter approach essentially presents thesignal to the user. Vibrotactile devices present informationthe user via mechanical depression of the skin, with the sbeing more sensitive to vibrations in the frequency rangeto 250 Hz~Sherrick and Craig, 1982!. Electrotactile devicespresent information via electrical stimulation of the sensreceptors or nerve fibers in the skin. This stimulation is mefficiently conducted using current pulses, with rates inrange of 20 to 400 pulses per second ensuring that individpulses are not felt but that the nerve has sufficient timerecover between pulses~Brown and Stevens, 1992!. The dis-advantages of vibrotactile stimulation are that relatively la

2971 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

al

d-ederet

s

yo-re

-

e-e--

is

5li

les-

f

of

theessof

awoin0

ytealo

e

transducers and a relatively high rate of power consumpare required~Sherrick, 1984!. The disadvantages of electrotactile stimulation are the greater potential for unpleasstimulation~Brown and Stevens, 1992!, and the limited dy-namic range of the tactile system for responding to variatin the intensity of electrotactile stimulation; this latter disavantage requires that sound input be compressed or intecoded in an alternative manner~Weisenberger, 1992!.

The Tickle Talker™ consists of a lapel microphone,handset, and a speech processor. The eight electrodes ifour rings of the handset are positioned over the digital nebundles, which lie along the sides of the fingers. Informatis presented to the user via electrocutaneous stimulatiothese nerve bundles. The original Tickle Talker™ speeprocessor was based on the WSP II processor used withoriginal version of the Nucleus 22-channel cochlear impl~Blamey and Clark, 1985!. The speech processor used in tpresent study was the Mk III version, which incorporatednumber of design improvements including the use ofnonactive finger electrodes as the ground/return electrduring stimulation. The speech-processing strategy impmented in this processor extracted estimates of the funmental frequency and the second formant frequencyspeech, and the overall amplitude of speech in the band u4 kHz and in a second band above 4 kHz~Cowan et al.,1991!. These features were encoded as electrical stimuparameters, withF0 being encoded as pulse rate,F2 beingencoded as electrode position on electrodes 2 to 7, oveamplitude being encoded as stimulus intensity, and hifrequency information resulting in stimulation on electro8. In addition to the frequency-specific stimulation on eletrodes 2 to 7, the voiced/unvoiced contrast was presentestimulation on electrode 1 for unvoiced sounds. No stimution was presented on electrode 1 for voiced sounds.crossover boundaries for the electrodes were 1100, 11500, 1800, 2000, 2200, and 4000 Hz.

C. Subjects

The subjects were two females~S1 and S5! and twomales~S2 and S3! from the group of six normally hearingadults who participated in a previous training study~Galvinet al., 1999!. Subject numbering is consistent with that studPersonal commitments prevented S4 and S6 from beingvolved in the present study. By coincidence, S4 and S6 donstrated the fastest and the slowest rates, respectivelword learning in the training study, so that the four remaing subjects form a relatively homogeneous group withspect to learning rate. The number of hours of tactile-aloword recognition training received, the number of worlearned, and the learning rate for each subject in the trainstudy are presented in Table II. The subjects had receivevisual-alone or combined-modality training, and no tactialone training other than the word-level training providedthe previous training study.

D. Experimental setup

The experimental setup was identical to that used fortraining study ~Galvin et al., 1999!. To ensure that the

2971Galvin et al.: Generalization of tactile perceptual skills

gsron

kee,tha

le

onioito

u.

onliasttthteantioer

tos

rnedsfoa

pnprltengn.t

rdrd

thanist

nali-p-h a

ds-achho-to

otere

ofuch

thechta-ed-ual-of

hessed

d forrds

sen-ereest,nded

ions

ol-ds;d byen-dayp-entit-

his

e

tence

ditio

stimuli were not audible, the subject wore foam earplureceived white noise via headphones, and was isolated fthe talker inside a soundproof booth. During visual-alone atactile–visual evaluations, the subject viewed the talthrough the booth window. A Mirage DM520 microphonpositioned approximately 15 cm from the talker’s moureplaced the standard Tickle Talker™ lapel microphone,lowing the input signal to be delivered through a mixer.

E. Evaluation

Details of the training and evaluation procedures impmented in the training study were reported in Galvinet al.~1999!. Briefly, the subjects received tactile-alone traininga closed set of words that increased in size as evaluatindicated that the vocabulary was being recognized w70%–80% accuracy. Within 10 days of the cessationtraining, the subjects were evaluated in the present stwith a variety of tactile-alone and tactile–visual materials

The generalization of tactile-alone word recognitiskills to unfamiliar talkers was assessed with an unfamimale talker and an unfamiliar female talker. A 34-word lion which all subjects had been trained and evaluated intraining study, was used. The evaluation protocol wassame as that used during the training study; that is, thelist consisted of two random presentations of each word,subjects responded from a closed-set list. Initial evaluawas with the familiar male talker, followed directly by thtwo other unfamiliar talkers. The order of these two talkewas randomized across subjects.

The generalization of tactile-alone perceptual skillsthe perception of words in sentences was assessed usingtences constructed only from the words each subject leaduring training. Details of the individual test list constructfor each subject are presented in Table III, and examplethe sentences used are presented as Appendix A. Two towords in each sentence were designated as key words,the sentences were scored by the number of key wordsceived correctly. The first subject evaluated, S2, had lear46 words. Construction of a sentence list based on thesentation of each learned word twice as a key word resuin 92 key words in 33 sentences. This test list was too lomaking it difficult for the subject to maintain concentratioFor subsequent subjects, the evaluation protocol used intraining study was followed, so that the number of key wowas limited to a maximum of 70, with each learned wo

TABLE II. Number of hours of training received, number of words learneand learning rate for subjects in a previous tactile-alone word recogntraining study.a

Subject Hours of trainingb Words learned Rate~words/h!

S1 33.1 77 2.3S2 17.8 46 2.6S3 12.3 34 2.8S4 17.7 27 1.5S5 18.5 62 3.4S6 12.5 52 4.2

aGalvin et al. ~1999!.bExcludes evaluation time.

2972 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

,mdr

,l-

-

nshf

dy

r,heestdn

s

en-ed

ofurnd

er-ede-d,

hes

being presented as a key word at least once but no moretwice. Note that subjects were not provided with a word lto refer to during the sentence testing.

The generalization of tactile-alone word recognitioskills to the perception of untrained words, and the generzation of tactile perceptual skills to the tactile–visual percetion of trained and untrained words, were assessed witsingle word list containing both word types. The list~at-tached as Appendix B! was constructed from 20 untrainewords and 20 trained words~these being the first 20 wordused in the training study!. The untrained words were constructed from the phonemes of the trained words, with ephoneme used only once and no consideration given to pneme position within a word. This procedure was usedachieve the maximum similarity between the lists. It is nclaimed that the procedure produced word lists that wphonemically balanced, as changes in the word positionthe phonemes clearly change their acoustic properties. Sbalancing would be impossible to achieve, given thatcontent and order of the trained word list was fixed. Eaadministration of the list consisted of two random presentions of each word, with subjects responding from a closset list. Subjects were evaluated in the tactile-alone, visalone, and tactile–visual conditions, with the orderconditions balanced across subjects.

The generalization of tactile perceptual skills to ttactile–visual perception of open-set material was asseusing the Consonant–Nucleus–Consonant~CNC! words test~Peterson and Lehiste, 1962! and the Speech IntelligibilityTest ~SIT! sentences~Magner, 1972!. For the CNC words,each list contained 50 words, and responses were scorethe number of phonemes correct and the number of wocorrect. For the SIT sentences, each list contained 16tences with a total of 80 key words, and responses wscored for the number of key words correct. For each ttwo lists were presented in each of the visual-alone atactile–visual conditions. The lists used were randomizacross subjects and conditions, and the order of conditwas balanced across subjects.

The evaluation materials were administered in the flowing order: sentences constructed from trained worclosed-set trained and untrained words; words presenteunfamiliar talkers; and the CNC words test and the SIT stences. Nearly all evaluations were completed in the 10-period following the cessation of training. The only excetion was that S3 completed the unfamiliar talker assessm1 month after the cessation of training, due to other commments of this subject and the unfamiliar male talker. T

TABLE III. Details of the test list for each subject for the tactile-alonperception of key words in sentences constructed from trained words.

Subj.Trained words used to

construct sentencesSentences/

listKey words/

listChancea

~%!Score~%!

S1 70 24 70 3.0 14.2S2 46 33 92 7.4 6.5S3 34 24 68 8.6 17.6S5 62 24 70 3.5 22.8

aCalculated by dividing the average number of words per response senby the number of trained words used to construct the sentences.

,n

2972Galvin et al.: Generalization of tactile perceptual skills

nswawaswhas

ithfieteine-e

ngthusf-um

wur

fodthic

rn-

thrdh

orre

s ofjected,theushadtorstwotedfa-

ith

nofa-p-re

oiceveltectthe

le-ned

bynceen-that

theun-nfi-lcu-nd

ithfeor

nd

ns the

deviation from the protocol did not affect the conclusiodrawn, as the most important comparison of scoresmade across talkers, and S3 completed the evaluationsall three talkers on the same day. For the CNC words testthe SIT sentences, the post-training evaluation was denated evaluation 2. For these tests only, evaluation 1conducted during the training study, after the subjectslearned 25 to 30 words. Evaluation 1 was defined a‘‘midtraining’’ evaluation.

The evaluation materials were presented live-voice wout feedback. With the exception of the evaluations specias involving unfamiliar talkers, all materials were presenby the same male talker who previously provided the traing. The familiar and the two unfamiliar talkers were rsearchers or clinicians experienced in producing clear spefor artificially deafened research subjects or heariimpaired clients. The talkers were instructed to presentevaluations using the clear speech they would normallyfor communicating with hearing-impaired or artificially deaened individuals. For the sentence evaluations, a medipaced rate of presentation was used so that the speechclear but the rate was not slowed to the extent that natintonation patterns were affected.

III. RESULTS

Figure 1 presents the individual and group scorestactile-alone word recognition with the familiar talker anthe two unfamiliar talkers. The dashed line represents95% confidence limit for scores greater than chance, whwas calculated using the binomial approximation of Thoton and Raffin~1978!. As shown, all subjects scored significantly above chance with the familiar talker and both ofunfamiliar talkers. When evaluated with the familiar talkeS1 and S5 scored above the 70% evaluation criterion usethe training study, whilst S2 and S3 scored just 57.4%. Tword recognition skills of S2 and S3 may have been maffected by the fact that training was no longer being

FIG. 1. Individual and group scores for tactile-alone word recognition wthe familiar male talker, the unfamiliar male talker, and the unfamiliarmale talker. The dashed line represents the 95% confidence limit for scgreater than chance@based on the binomial approximation of Thornton aRaffin ~1978!#.

2973 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

sithndig-asda

-dd-

ch-ee

-as

al

r

eh-

e,inee-

ceived when these evaluations were completed. The skillS3 may have been particularly affected due to this subcompleting this evaluation 4 weeks after training ceaswhile the other subjects completed this evaluation within10-day post-training period. In addition, in the previotraining study, S1 had received more training, and S5demonstrated a faster rate of word learning, and these facmay have contributed to the superior scores of thesesubjects. The chi-square test for goodness of fit indicathat, for each subject and the group, the scores with themiliar talker were significantly higher than the scores wthe unfamiliar male talker@x2(1,n568).10.8, p,0.001#and the unfamiliar female talker@x2(1,n568).4.2, p,0.04#. The chi-square test also indicated that there wassignificant difference between the scores with the two unmiliar talkers for S1, S2, S5, or the grou@x2(1,n568),1.9, p.0.19#. For S3, the score with the unfamiliar male talker was significantly lower than the scowith the unfamiliar female talker~x259.3, p50.002!. Itshould be noted that S3 reported after the event that the vof the unfamiliar male talker was presented at a lower lethan those of the other talkers and had been difficult to deat times. A possible equipment malfunction may explainpoor score of this subject with this unfamiliar talker.

Table III presents the key-word scores for the tactialone perception of sentences constructed from traiwords. Chance scores for each subject were calculateddividing the average number of words per response senteby the number of trained words used to construct the stences. The chi-square test for goodness of fit indicatedscores were significantly above chance for S1@x2(1,n570)531.3,p,0.001#, S3@x2(1,n568)57.3,p50.007#, and S5@x2(1,n570)577.7, p,0.001#, but not for S2 @x2(1,n592)50.1, p50.75#.

Figure 2 presents the individual and group scores fortactile-alone recognition of a closed set of trained andtrained words. The dashed line represents the 95% codence limit for scores greater than chance, which was calated using the binomial approximation of Thornton a

-es

FIG. 2. Individual and group (n54) scores for the tactile-alone recognitioof closed-set trained and untrained words. The dashed line represent95% confidence limit for scores greater than chance@based on the binomialapproximation of Thornton and Raffin~1978!#.

2973Galvin et al.: Generalization of tactile perceptual skills

lyorein

helir

cts

e5igub

pfed

neyp

cer

riedthe

t-con-asn-

ifi-sub-ceile–ionereor

n the1

ndly-

ging

u

ned-andace

ITl con-

Raffin ~1978!. As shown, all subjects scored significantabove chance on trained and untrained words. The grscore of 42.5% on trained words was lower than thatported for the 34-word list in Fig. 1 and the criterion usedthe training study. This difference is most likely due to tfact that the 20 trained words were presented in a singlethat also included 20 untrained words. An analysis of vaance~ANOVA ! was conducted on the scores of all subjefor trained and untrained words. The factors in the analywere subject and word type~trained or untrained!. The maineffect of word type was significant@F(1,3)514.2, p50.033#, indicating that the mean score of 42.5% on trainwords was significantly higher than the mean score of 22.on untrained words. The main effect of subject was not snificant, indicating that performance was similar across sjects.

Figures 3~a! and ~b! present the individual and grouscores for the visual-alone and tactile–visual recognition oclosed set of trained words and untrained words, resptively. An ANOVA was conducted on the visual-alone antactile–visual scores of all subjects for trained and untraiwords. The factors in the analysis were subject, word t~trained or untrained!, and condition. The resultingF statis-tics andp values are presented in Table IV. The main effeof subject was significant, indicating that, when scores w

FIG. 3. Individual and group scores for the visual-alone and tactile–visrecognition of closed-set trained words~a! and untrained words~b!.

2974 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

up-

sti-sis

d%--

ac-

de

te

collapsed across conditions and word type, the scores vasignificantly across subjects. Tukey’s test indicated thatmean score of S3 was significantly lower (p50.04) than thatof S1. The main effect of word type was significant, indicaing that, when scores were collapsed across subjects andditions, the means score of 78.4% for trained words wsignificantly higher than the mean score of 63.1% for utrained words. The main effect of condition was not signcant, indicating that, when scores were collapsed acrossjects and word type, there was no significant differenbetween the mean scores in the visual-alone and tactvisual conditions. The absence of any significant interacteffects indicated that the findings for the main effects wnot significantly different across subjects, conditions,word types.

Figure 4 presents the mean group scores~collapsedacross two lists and four subjects! for CNC words, CNCphonemes, and SIT sentences. Scores were obtained ivisual-alone and tactile–visual conditions at evaluation~midtraining! and evaluation 2~post-training!. For each ma-terial, an ANOVA was conducted on the visual-alone atactile–visual scores of all subjects. The factors in the anasis were subject, evaluation, and condition. The resultinFstatistics andp values are presented in Table V. The maeffect of subject was significant for all materials, indicatin

al

TABLE IV. F statistics andp values for the main effects and interactioeffects of the ANOVA conducted on the scores for the perception of closset words. Scores were obtained from the four subjects for traineduntrained words in the visual-alone and tactile–visual conditions. Boldftype indicates effects which reached significance ata50.05.

Effect F p

Subjecta 9.8 0.047Word typeb 40.2 0.008Conditionb 1.4 0.33Cond3word typeb 2.0 0.25Cond3subja 3.6 0.16Word type3subja 1.7 0.34

aDegrees of freedom5~3,3!.bDegrees of freedom5~1,3!.

FIG. 4. Group (n54) scores for CNC words, CNC phonemes, and Ssentences. Scores were obtained in the visual-alone and tactile–visuaditions at evaluation 1 and evaluation 2.

2974Galvin et al.: Generalization of tactile perceptual skills

anth

he(e

aitct

nju%ifierelyo

onececuon

hendnt

tualeea

ereat-isinthee–inof

rds,rdsged

nceeenionotre-edh ath-

ofd in

i-

e toents-

on,-

atheythealer-

an

thet inw-

, ora

byu-

eener’s

hasto

on-illsheril-to

nhoctndt

es

that, when scores were collapsed across conditionsevaluations, the scores of some subjects were significadifferent from some others. Tukey’s test indicated thatmean scores of S1 were significantly higher (p,0.002) thanthe scores of the other subjects for each material. Furtmore, the mean score of S3 was significantly lowerp,0.008) than those of all other subjects for SIT sentencand lower than that of S2 for CNC phonemes. The meffect of evaluation was significant for all materials, indicaing that, when scores were collapsed across subjects andditions, the mean scores at evaluation 2 were higher thanmean scores at evaluation 1. However, the actual differein scores at the two evaluations was quite small, being3.4% for CNC phonemes, 5.3% for CNC words, and 4.6for SIT sentences. The main effect of condition was signcant for CNC phonemes, indicating that, when scores wcollapsed across subjects and evaluations, the mean sco47.2% in the tactile–visual condition was significanthigher than the mean score of 44.2% in the visual-alone cdition. The only significant interaction effect was a conditiby evaluation interaction effect for SIT sentences. This effindicated that, when scores were collapsed across subjthe mean score at evaluation 2 of 39% in the tactile–viscondition was significantly higher than the mean score27.3% in the visual-alone condition. The absence of aother significant interaction effects indicated that, with texception of the condition effect for SIT sentences, the fiings for the main effects were not significantly differeacross subjects, evaluations, or conditions.

IV. DISCUSSION

The results demonstrate that the tactile-alone percepskills learned during word recognition training were generized to new tactile-alone contexts, with subjects able to rognize trained words presented by unfamiliar talkers ofther gender, trained words presented in sentences,

TABLE V. F statistics andp values for the main effects and interactioeffects of the ANOVAs conducted on scores for CNC words, CNC pnemes, and SIT sentences. Scores were obtained from the four subjeevaluation 1 and evaluation 2 in the visual-alone and tactile–visual cotions. Boldface type indicates effects which reached significance aa50.05.

Effect CNC words CNC phonemes SIT sentenc

Subjecta F 18.0 44.8 164.9p Ë0.001 Ë0.001 Ë0.001

Evaluationb F 7.4 7.4 7.2p 0.015 0.015 0.017

Conditionb F 1.1 5.8 3.3p 0.32 0.028 0.089

Cond3evalb F 0.1 0.04 17.0p 0.8 0.84 0.001

Cond3subja F 2.7 1.6 2.4p 0.08 0.23 0.11

Eval3subja F 1.4 2.7 2.3p 0.3 0.082 0.12

Eval3subj3conda F 1.0 0.6 2.3p 0.42 0.66 0.12

aDegrees of freedom5~3,16!.bDegrees of freedom5~1,16!.

2975 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

ndtlye

r-

s,n-on-hecest

-re

of

n-

tts,

alfy

-

al-c-i-nd

tactually new untrained words. The scores achieved wlower than the evaluation criterion used in training, indicing that complete generalization of skills did not occur. Thwas not surprising, given that the training was limitedamount, type, and number of trainers. In contrast totactile-alone results, the generalization of skills to tactilvisual contexts was very limited, with no improvementtactile–visual over visual-alone scores for the perceptionclosed-set trained and untrained words or open-set woand minimal improvement for phonemes in open-set woand words in open-set sentences. It should be acknowledat this point that it is possible that the subjects’ above-chaperformance on the tactile-alone evaluations may have bdue to their baseline capacity for tactile-alone perceptwith the Tickle Talker™; i.e., their performance was ninfluenced by the tactile-alone word recognition trainingceived. However, the extensive clinical experience gainwith the device by the authors strongly suggests that suclevel of tactile-alone perception could not be achieved wiout experience and/or training, and that the performancethe present subjects was a function of the training receivethe previous study.

Clinical experience with the Tickle Talker™ and prevous evaluations of vowel and consonant recognition~Cowan,1991! suggested that tactile perceptual skills do generalizunfamiliar talkers, and this was confirmed by the presscores for unfamiliar talkers. However, talkers vary in apects of production such as pronunciation, vowel durativowel F2, and frication~Peterson and Barney, 1952; Nabelek, Czyzewski, and Krishnan, 1992!. The lower scores forunfamiliar as compared with familiar talkers indicates thsome tactile cues varied sufficiently between talkers that twere not recognized when an unfamiliar talker presentedwords. The additional variation between male and femtalkers did not appear to affect word recognition perfomance. In particular, female talkers have a higherF2 fre-quency than males, with the median difference for Americtalkers reported as 250 Hz~Peterson and Barney, 1952!. De-pending upon the frequency of a particularF2 in relation tothe frequency boundaries of the electrodes, a change inF2 frequency with a change in talker gender could resulthe movement of stimulation to an adjacent electrode. Hoever, it appears that this either did not occur consistentlydid affect word recognition. Female talkers also havehigherF0 than males. Prosodic information is conveyedthe Tickle Talker™ as relative changes in the rate of stimlation. The perception of relative changes may have bunaffected by the increase inF0 with the change to a femaltalker. It is also possible that, irrespective of the talkegender, the subjects were just not using theF0 informationprovided by the Tickle Talker™, as previous researchindicated that stimulus rate is the most difficult parameterperceive in the tactile signal~Blamey and Clark, 1987!.

In the study of Brooks and colleagues~Scilley, 1980!,the two subjects trained with female talkers also demstrated generalization of tactile-alone word recognition skto unfamiliar talkers that was independent of gender. Furtinformation regarding the generalization of skills to unfamiar talkers would be gained through using a new talker

-s ati-

2975Galvin et al.: Generalization of tactile perceptual skills

uin

ucrouma

lizonh

ev

he

orrd

an-

piintolndded

oa. Iofu

jenuo

cei

chbein

linanhtionng

tu

labyth

nom

andt re-ed43

rromosethatthesks,uf-e-

ject

thatfor-

ub-thentherds

therds.dno-

outects’as

therds

tilext oftoomedur-

ep-/

d asrdith-rdstheun-

%

%0

set

–tedotnd

ous

repeat the training study with the present subject groGiven the results so far, an accelerated rate of word learncould be expected during training with a second talker. Sa result was reported for one subject of Brooks and F~1983!, who demonstrated a continuing decrease in the nber of training sessions required to learn 100 words as eof four unfamiliar talkers was introduced.

Three of the present subjects demonstrated generation of tactile-alone word recognition skills to the perceptiof key words in sentences, with a mean score of 15.3%. Tscore was higher than the 1% tactile-alone score of a prous group evaluated on BKB~Bamford–Kowal–Bench! sen-tences following visual and tactile–visual training using tTickle Talker™ ~Blamey et al., 1989!, but lower than thescores generally demonstrated in the present study for win isolation. Compared to the presentation of isolated wothe sentence format provides additional syntactic and semtic information ~Boothroyd and Nittrouer, 1988; Olsen, VaTasell, and Speaks, 1997!, so that subjects performing reasonably well in auditory or visual speech perception tycally obtain higher scores for the recognition of wordssentences. In the present study, the predictability of mosthe sentences was low, and subjects usually recognizedone or two words per sentence. Therefore, the potentiautilize semantic and syntactic information was limited, asubjects were most likely to be attempting to recognize invidual words within the sentences. The lower scores wlikely due to the difficulty of perceiving word boundaries anrecognizing words whose tactile pattern was altered by cticulatory effects. Memory may also have played a rolesubjects made perceptual decisions at the completionstimulus, then the tactile pattern of the whole sentence woneed to have been retained. If, on the other hand, submade decisions during the stimulus presentation, then othe recognized words would need to be retained. As the sjects were not queried on this point, it is not known whichthese scenarios occurred.

This study does not suggest that the tactile-alone pertion of conversational-level speech could be achievedpractice with the Tickle Talker™ at normal rates of speeand provides only minimal evidence that this couldachieved at a slow, word-by-word rate. Tactile-alone runnspeech perception has only been reported for usersTadoma, a communication method developed for deaf–bpeople to obtain tactile speech information by resting a hacross the face and throat of a talker. Tadoma was taugthe mid-twentieth century as the primary communicatmethod in a small number of American schools for heariimpaired children~Schultzet al., 1984!. A major social dis-advantage of the method was the requirement of touchingtalker. Speechtracking scores of 30 to 40 words per minhave been reported for experienced Tadoma users~Reedet al., 1985!. Such success has been attributed, in particuto the rich multidimensional articulatory display providedthe face and the extensive training and experience ofsubjects~Reedet al., 1989, 1992!. It is impractical for sub-jects in tactile device studies to obtain levels of training aexperience comparable with those of experienced Tadusers. As a consequence, valid comparisons can only

2976 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

p.ghst-

ch

a-

isi-

dss,n-

-

ofnlyto

i-re

r-fa

ldctslyb-f

p-n,

gofddin

-

hete

r,

e

da

be

made between the performance of tactile device usersthe results reported for naive Tadoma users. The presensults can be compared with those of two inexperiencTadoma users who received 26 h of training and learnedwords ~Reedet al., 1982!. Average post-training scores fothe tactile-alone perception of sentences constructed ftrained words were 40% and 20%. Superior scores to thof the present subjects were to be expected, givenReed’s subjects received more training than most ofpresent subjects, their training included sentence-level taand they were using Tadoma, which is known to supply sficient information for running speech perception. Neverthless, with a score of 20%, Reed’s poorer performing subscored similarly to S3~17.6%! and S5~22.8%! of the presentgroup. Although interesting, this result does not suggestexperienced Tickle Talker™ users could emulate the permance of experienced Tadoma users.

In our previous training study, it was assumed that sjects were learning to perceive tactile speech cues andusing these skills to recognize new words, given thatsubject group required less training time to learn new woas the study progressed~Galvin et al., 1999!. The scores inthe present study for untrained word perception supportassumption that skills were generalized to untrained woLynch et al. ~1988! also concluded from post-training worrecognition scores that their subjects were applying phologically based tactile knowledge. Brookset al. ~1987! con-cluded that subjects were able to apply general rules abthe tactile representation of speech features, as their subjpost-training score for consonant feature recognition waround 87%.

The incomplete generalization demonstrated byhigher scores on trained as compared with untrained wowas possibly due to one or more of four factors. First, tacspeech cues may have been easier to detect in the contea trained word as this particular context would be familiarthe subject. Second, the subtle differences between swords may have become clearer with repeated exposureing training. Third, the subjects reported that the tactile rresentation of some words~for example, words containing /sor /b/ or long vowels! was highly distinctive so that, withrepeated exposure, these words were easily recognizewhole words. Finally, the subjects reported that if a wowas recognized as a trained word they often responded wout consulting the response list, where as for untrained wothe memory of the tactile pattern had to be retained whilelist was scanned. Previous studies have also evaluatedtrained word recognition, with reported scores of 4% to 5for two adults using the Tacticon 1600~Lynch et al., 1988!,4% for four children using the Tacticon 1600, around 30for four children using the Tactaid II or the Tacticon 160and selecting from a limited response set~Lynch et al.,1989!, and 5.1% for one subject evaluated with open-words ~Brookset al., 1986a!.

The generalization of tactile-alone skills to tactilevisual perception was very limited. This was an unexpecresult, given that the tactile signal provided information neasily available visually, such as final consonant voicing avowel second formant. On the other hand, some previ

2976Galvin et al.: Generalization of tactile perceptual skills

sthth

ion

-seri

intil

te2

dsti-sso

e

don

nr™miondyniv

ts

ngteea

ryun–

g.ern,t

ea

ne

odneiil

ed

leum-theedro-tileng-

nd,nots atig-

nalweroorion

easa-

vel.in-nden-a-

theheave

theg-asthein-

h as

etile-t the

m-di-ingednewforvi-fulheor-opful

studies with the Tickle Talker™~Blamey et al., 1989! andwith other tactile devices@see a review by Bernstein~1992!#have demonstrated that subjects may be able to obtainnificant amounts of tactile speech-feature information wiout necessarily gaining an equivalent level of benefit incombined-modality condition.

These results for combined-modality perception arecontrast to those reported in other studies of tactile-alword recognition training. Brookset al. ~1986a, 1986b! re-ported percentage-point improvements of 29.3% on openwords and 17% on open-set sentences when the QueUniversity vocoder was added to vision. The very expeenced subject had received 65 h of tactile-alone trainLynch and colleagues compared tactile-alone and tacauditory performance. Lynchet al. ~1989! reported that chil-dren using the Tacticon 1600 or the Tactaid II demonstrarespective percentage-point improvements of 39% and 2for trained words, and 27% and 8% for untrained worLynch et al. ~1988! reported that two adults using the Taccon 1600 demonstrated percentage-point improvement24% for trained words and 15% for untrained wordLynch’s studies differed from the present study in the usethe tactile–auditory evaluation condition, hearing-impairsubjects, and pediatric subjects who had months of multimdality experience using the device in the classroom. In adtion, the adults received an average of 44 h of tactile-altraining.

The results of previous studies examining tactile-aloand combined-modality perception using the Tickle Talkemay contribute to understanding the present results. Blaet al. ~1989! compared observed and predicted informattransmission scores for combined modalities, and concluthat combining tactile information with visual or auditorinformation was less effective than combining auditory avisual information. The authors suggested that the effectness of the auditory and visual combination might be dueexperience and/or the presence of neural mechanismscialized to this task. Alca´ntaraet al. ~1993! provided unimo-dal or tactile–auditory speech-feature perception trainiThe unimodally trained subjects demonstrated the greapre- to post-training improvements in tactile–auditory vowand consonant recognition, and combined the tactile andditory information as effectively as the tactile–auditotrained subjects. The authors concluded that improvedmodal perception would contribute to improved tactileauditory perception without combined-modality traininHowever, it is important to note that these conclusions wlimited to analytic training, the tactile–auditory conditioand speech-feature perception~as similar results were noachieved for the perception of phonemes in words!. In aspeech-feature training study by Oerlemans and Blam~1998!, subjects received tactile–visual training or unimodtraining or tactile–visual training whilst concentrating on ospecified modality~bimodal-cued!. A greater pre- to post-training increase in scores was demonstrated by the bimgroup for the perception of the vowel, but not consonafeatures in words. The authors proposed that the complemtary nature of the tactile and visual modalities resultedgreater benefits being gained from bimodal training, wh

2977 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

ig--e

ne

etn’s-g.e-

d%.

of.f

do-i-e

e

ey

ed

de-ope-

.stlu-

i-

e

yl

alt,n-

ne

the redundancy of the auditory and tactile information usin the earlier study of Alca´ntara et al. ~1993! resulted innothing ‘‘extra’’ to be gained from bimodal training.

In conclusion, these previous studies with the TickTalker™ and those with other tactile devices suggest a nber of factors relating to the tactile device, the subjects,training, and the evaluation modality may have influencthe results obtained. First, the Tickle Talker™ may have pvided less useful speech information than the other tacdevices, although scores reported previously for heariimpaired adults~Alcantara, 1991; Cowanet al., 1988! sug-gest it is unlikely that this is the sole explanation. Secothe normally hearing subjects of the present study mayhave been as skilled as other hearing-impaired subjectintegrating incomplete speech signals and/or may havenored the tactile signal when the more familiar visual sigwas available. It has previously been suggested that lothan predicted tactile–visual scores may be due to a pstrategy being employed in combined-modality percept~Summers and Du, 1997!. Third, the use of ‘‘higher level’’training materials~i.e., words! may not have been effectivin developing combined-modality perceptual skills if, as hpreviously been suggested for auditory and visual informtion ~Blamey, 1990; Massaro, 1987!, visual and tactile infor-mation are combined at or before the speech-feature leFourth, the amount of training provided may have beensufficient to achieve improved perception at the word asentence level. Finally, the use of the relatively complemtary tactile and visual conditions in the post-training evalutions may have resulted in no benefit being shown fromunimodal training. It is not possible to determine from tpresent study how influential each of these factors may hbeen.

The results of this study can be used to considerpotential benefits of tactile-alone training for hearinimpaired users of the Tickle Talker™. As the device wdesigned to provide supplementary speech information,usual training program emphasizes combined-modality traing, and only includes some basic tactile-alone tasks, sucsound detection and phoneme discrimination~Galvin et al.,1993!. Given the information provided by the device, thresults of previous speech perception studies, and the tacalone sentence scores in the present study, it is clear tharole of the Tickle Talker™ will remain that of providingsupplementary speech information. The failure here to deonstrate generalization of skills to the tactile–visual contion suggests that extensive tactile-alone, word-level trainwould not achieve maximum benefits for hearing-impairdevice users. Nevertheless, the generalization of skills totactile-alone contexts indicates that there may be a rolemore tactile-alone training than has been included preously. Tactile-alone training at the word level may be usefor establishing tactile perceptual skills and introducing tdevice to the user, and will definitely provide a good opptunity to focus only on the tactile information and to develthe understanding that the tactile signal provides usespeech information.

2977Galvin et al.: Generalization of tactile perceptual skills

sheaao

.thm

nculngaintdylithth

ce

llsldm

telio

rouo

smTa

d,eremrcinraaand

tio

ctt

–20

inge

ithn to

-

ring

ooc.

al

a-

fi-Dev.

en’s

V. CONCLUSION

The tactile-perceptual skills learned by these subjecttactile-alone word recognition training generalized to ttactile-alone perception of sentences, untrained words,words presented by unfamiliar talkers. There was minimevidence of generalization to the tactile–visual perceptionclosed-set words, open-set words, or open-set sentencestactile-alone results suggest that tactile-alone training atword level may be a useful component of a training prograhowever, the tactile–visual results reinforce the importaof combined-modality training. It appears from this resthat the emphasis of tactile training programs for heariimpaired device users should be on combined-modality tring. However, future research needs to consider this poinmore detail for, unlike the subjects of the present stuhearing-impaired device users will gain combined-modaexperience with a tactile device in daily communication. Tquestion that remains unanswered by this study is whehearing-impaired device users needformal combined-modality training, or whether combined-modality experiengained in daily device use may be equally effective in devoping the important combined-modality perceptual skiClearly, if daily experience were equally effective it woube possible to spend more of the usually finite training tion learning to detect and discriminate the speech cues intactile signal. It is possible that this may be most effectivachieved through tactile-alone training. Further investigatof this issue is required.

ACKNOWLEDGMENTS

The authors are grateful for the technical support pvided by Anne Lu, and the contributions of three anonymoreviewers. The study was funded by the CommonwealthAustralia, Department of Industry, Science and Tourithrough the Cooperative Research Centres Program.contribution of the parties to the then CRC for CochleImplant, Speech and Hearing Research is acknowledgeparticular those of The Bionic Ear Institute and The Univsity of Melbourne’s Department of Otolaryngology. ThTickle Talker™ was developed with financial support frothe then Department of Industry, Technology and Commeand the Department of Employment, Education and Trainof the Commonwealth Government of Australia, the Austlian Research Council through the funding of the HumCommunication Research Centre, the National HealthMedical Research Council of Australia, Cochlear Limitethe George Hicks Foundation, and the Ian Potter Founda

APPENDIX A

Examples of sentences used in the assessment of taalone sentence perception. Key words used for scoringsubject’s response are underlined.

~1! Put thepaper under thehouse.~2! The best jam isred.~3! The baby doesn’t drinkmuch.~4! Did they see theplane?~5! The music show isover.

2978 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

in

ndlf

Thee;et--

in,

yeer

el-.

eheyn

-sf

herin

-

eg-nd

,n.

ile-he

APPENDIX B

Word list for the tactile-alone, visual-alone, and tactilevisual perception of trained and untrained words. Thetrained words were the first 20 words used in the trainstudy ~Galvin et al., 1999!. The 20 untrained words werconstructed from the phonemes of the 20 trained words, weach phoneme used only once and no consideration givephoneme position within a word.

Trained words Untrained words

~1! boy ~1! stork~2! put ~2! peel~3! we ~3! wood~4! see ~4! grave~5! door ~5! wing~6! they ~6! geese~7! house ~7! shower~8! show ~8! goat~9! go ~9! thieve~10! thing ~10! soy~11! what ~11! though~12! sing ~12! ping~13! speak ~13! thongs~14! talking ~14! swung~15! something ~15! soap~16! radio ~16! wart~17! giving ~17! bake~18! over ~18! doing~19! little ~19! later~20! paper ~20! theme

Alcantara, J. I.~1991!. ‘‘The effect of training on tactile speech perception,’’ Unpublished Doctoral thesis~University of Melbourne, Australia!.

Alcantara, J. I., Blamey, P. J., and Clark, G. M.~1993!. ‘‘Tactile–auditoryspeech perception by unimodally and bimodally trained normal heasubjects,’’ J. Am. Acad. Audiol.4, 98–108.

Bernstein, L. E.~1992!. ‘‘The evaluation of tactile aids,’’ inTactile Aids forthe Hearing Impaired, edited by I. R. Summers~Whurr, London!.

Blamey, P. J.~1990!. ‘‘Multimodal stimulation for speech perception,’’ inInformation Processing in Mammalian Auditory and Tactile Systems, ed-ited by M. Rowe and L. Aitkin~Wiley-Liss, New York!.

Blamey, P. J., and Clark, G. M.~1985!. ‘‘A wearable multiple-electrodespeech processor for the profoundly deaf,’’ J. Acoust. Soc. Am.77, 1619–1620.

Blamey, P. J., and Clark, G. M.~1987!. ‘‘Psychophysical studies relevant tthe design of a digital electrotactile speech processor,’’ J. Acoust. SAm. 82, 116–125.

Blamey, P. J., Cowan, R. S. C., Alca´ntara, J. I., Whitford, L. A., and Clark,G. M. ~1989!. ‘‘Speech perception using combinations of auditory, visuand tactile information,’’ J. Rehabil. Res. Dev.26, 15–24.

Boothroyd, A., and Nittrouer, S.~1988!. ‘‘Mathematical treatment of con-text effects in phoneme and word recognition,’’ J. Acoust. Soc. Am.84,101–114.

Brooks, P. L., Frost, B. J., Mason, J. L., and Gibson, D. M.~1986a!. ‘‘Con-tinuing evaluation of the Queen’s University tactile vocoder. I. Identifiction of open set words,’’ J. Rehabil. Res. Dev.23, 119–128.

Brooks, P. L., Frost, B. J., Mason, J. L., and Gibson, D. M.~1986b!. ‘‘Con-tinuing evaluation of the Queen’s University tactile vocoder. II. Identication of open set sentences and tracking narrative,’’ J. Rehabil. Res.23, 129–138.

Brooks, P. L., Frost, B. J., Mason, J. L., and Gibson, D. M.~1987!. ‘‘Wordand feature identification by profoundly deaf teenagers using the QueUniversity tactile vocoder,’’ J. Speech Hear. Res.30, 137–141.

Brooks, P. L., and Frost, B. J.~1983!. ‘‘Evaluation of a tactile vocoder forword recognition,’’ J. Acoust. Soc. Am.74, 34–39.

Brown, B. H., and Stevens, J. C.~1992!. ‘‘Electrical stimulation of the

2978Galvin et al.: Generalization of tactile perceptual skills

tile

es

k,troSo

chne

. MnsAm

. Mith

, Gg

ns

e

,’’

v

’’

for

nd

’

g

ntz,ar-

leus

itn-

ic

r.

g:

x-

nt,is-

y-d

L.

gh

nd

-

af

ev.

il-

f

t

’

c.

skin,’’ in Tactile Aids for the Hearing Impaired, edited by I. R. Summers~Whurr, London!.

Cowan, R. S. C.~1991!. ‘‘Speech feature encoding through an electrotacspeech processor,’’ Unpublished Doctoral thesis~University of Mel-bourne, Australia!.

Cowan, R. S. C., Alca´ntara, J. I., Blamey, P. J., and Clark, G. M.~1988!.‘‘Preliminary evaluation of a multichannel electrotactile speech procsor,’’ J. Acoust. Soc. Am.83, 2328–2338.

Cowan, R. S. C., Alca´ntara, J. I., Whitford, L. A., Blamey, P. J., and ClarG. M. ~1989!. ‘‘Speech perception studies using a multichannel electactile speech processor, residual hearing, and lipreading,’’ J. Acoust.Am. 85, 2593–2607.

Cowan, R. S. C., Blamey, P. J., Galvin, K. L., Sarant, J. Z., Alca´ntara, J. I.,and Clark, G. M.~1990!. ‘‘Perception of sentences, words, and speefeatures by profoundly hearing-impaired children using a multichanelectrotactile speech processor,’’ J. Acoust. Soc. Am.88, 1374–1384.

Cowan, R. S. C., Blamey, P. J., Sarant, J. Z., Galvin, K. L., and Clark, G~1991!. ‘‘Perception of multiple electrode stimulus patterns: Implicatiofor the design of an electrotactile speech processor,’’ J. Acoust. Soc.89, 360–368.

Engelmann, S., and Rosov, R.~1975!. ‘‘Tactual hearing experiment withdeaf and hearing subjects,’’ Except. Child.41, 243–253.

Galvin, K. L., Cowan, R. S. C., Sarant, J. Z., Blamey, P. J., and Clark, G~1993!. ‘‘Factors in the development of a training program for use wtactile devices,’’ Ear Hear.14, 118–127.

Galvin, K. L., Blamey, P. J., Oerlemans, M., Cowan, R. S. C., and ClarkM. ~1999!. ‘‘Acquisition of a tactile-alone vocabulary by normally hearinusers of the Tickle Talker™,’’ J. Acoust. Soc. Am.106, 1084–1089.

Geers, A. E., and Moog, J. S.~1995!. ‘‘Effectiveness of cochlear implantsand tactile aids for deaf children: The sensory aids study at Central Itute for the Deaf,’’ Volta Rev.95„5….

Kessler, D. K.~1993!. ‘‘The CLARION multi-strategy cochlear implant,’’Ann. Otol. Rhinol. Laryngol.177, 8–16.

Kishon-Rabin, L., Boothroyd, A., and Hanin, L.~1996!. ‘‘Speechreadingenhancement: A comparison of spatial–tactile display of voice fundamtal frequency (F0) with auditoryF0,’’ J. Acoust. Soc. Am.100, 593–602.

Lane, H., and Bahan, B.~1998!. ‘‘Ethics of cochlear implantation in youngchildren: a review and reply from a Deaf-World perspectiveOtolaryngol.-Head Neck Surg.119, 297–313.

Lynch, M. P., Eilers, R. E., Oller, D. K., and LaVoie, L.~1988!. ‘‘Speechperception by congenitally deaf subjects using an electrocutaneouscoder,’’ J. Rehabil. Res. Dev.25, 41–50.

Lynch, M. P., Eilers, R. E., Oller, D. K., and Cobo-Lewis, A.~1989!. ‘‘Mul-tisensory speech perception by profoundly hearing-impaired children,Speech Hear. Dis.54, 57–67.

Magner, M. E. ~1972!. A Speech Intelligibility Test for Deaf Children~Clarke School for the Deaf, Northampton, MA!.

Massaro, D. W.~1987!. Speech Perception by Ear and Eye: A ParadigmPsychological Inquiry~Erlbaum, Hillsdale, NJ!.

Nabelek, A. K., Czyzewski, Z., and Krishnan, L. A.~1992!. ‘‘The influenceof talker differences on vowel identification by normal-hearing ahearing-impaired listeners,’’ J. Acoust. Soc. Am.92, 1228–1246.

Oerlemans, M., and Blamey, P. J.~1998!. ‘‘Touch and auditory–visualspeech perception,’’ inHearing By Eye II, edited by R. Campbell, B.Dodd, and D. Burnham~Psychology, Hove, UK!.

Oller, D. K. ~1995!. ‘‘Tactile aids for the hearing impaired: An overview,’Semin. Hearing16, 289–295.

Olsen, W. O., Van Tasell, D. J., and Speaks, C. E.~1997!. ‘‘Phonemerecognition for words in isolation and in sentences,’’ Ear Hear.18, 175–186.

Osberger, M. J., Fisher, L., and Murad, C.~1996!. ‘‘Clinical results with theCLARION® multi-strategy cochlear implant in children,’’Cochlear Im-plants, Proceedings of the XVI World Congress of OtorhinolaryngoloHead and Neck Surgery, Sydney, 2–7 March 1997, edited by G. M. Clark~Monduzzi Editore, Bologna, Italy!.

2979 J. Acoust. Soc. Am., Vol. 108, No. 6, December 2000

-

-c.

l

.

.

.

.

ti-

n-

o-

J.

y

Parkinson, A. J., Parkinson, W. S., Tyler, R. S., Lowder, M. W., and GaB. J. ~1998!. ‘‘Speech perception performance in experienced cochleimplant patients receiving the SPEAK processing strategy in the NucSpectra-22 cochlear implant,’’ J. Speech Lang. Hear. Res.41, 1073–1087.

Peterson, G. E., and Barney, H. L.~1952!. ‘‘Control methods used in a studyof the vowels,’’ J. Acoust. Soc. Am.24, 585–597.

Peterson, G. E., and Lehiste, I.~1962!. ‘‘Revised CNC lists for auditorytests,’’ J. Speech Hear. Res.27, 62–70.

Plant, G.~1992!. ‘‘The use of tactile aids with congenitally deaf adults. Isworthwhile? A case study,’’Proceedings of the Second International Coference on Tactile Aids, Hearing Aids, and Cochlear Implants, edited byA. Risberg, S. Felicetti, G. Plant, and K.-E. Spens@Royal Institute ofTechnology ~KTH!, Department of Speech Communication & MusAcoustics, Stockholm, Sweden#.

Plant, G. L.~1995!. ‘‘Training approaches with tactile aids,’’ Semin. Hea16, 394–403.

Plant, G.~1998!. ‘‘Training in the use of a tactile supplement to lipreadinA long-term case study,’’ Ear Hear.19, 394–406.

Reed, C. M., and Delhorne, L. A.~1995!. ‘‘Current results of a field study ofadult users of tactile aids,’’ Semin. Hear.16, 305–315.

Reed, C. M., Doherty, M. J., Braida, L. D., and Durlach, N. I.~1982!.‘‘Analytic study of the Tadoma method: Further experiments with ineperienced observers,’’ J. Speech Hear. Res.25, 216–223.

Reed, C. M., Durlach, N. I., Delhorne, L. A., Rabinowitz, W. M., and GraK. W. ~1989!. ‘‘Research on tactual communication of speech: Ideas,sues, and findings,’’ Volta Rev. Mono.91, 65–78.

Reed, C. M., Rabinowitz, W. M., Durlach, N. I., Braida, L. D., ConwaFithian, S., and Schultz, M. C.~1985!. ‘‘Research on the Tadoma methoof speech communication,’’ J. Acoust. Soc. Am.77, 247–257.

Reed, C. M., Rabinowitz, W. M., Durlach, N. I., Delhorne, L. A., Braida,D., Pemberton, J. C., Mulcahey, B. D., and Washington, D. L.~1992!.‘‘Analytic study of the Tadoma method: Improving performance throuthe use of supplementary tactual displays,’’ J. Speech Hear. Res.35, 450–465.

Schultz, M. C., Norton, S. J., Conway-Fithian, S., and Reed, C. M.~1984!.‘‘A survey of the use of the Tadoma method in the United States aCanada,’’ Volta Rev.86, 282–292.

Scilley, P. L. ~1980!. ‘‘Evaluation of a vibrotactile auditory prosthetic device for the profoundly deaf,’’ Unpublished Masters thesis~Queen’s Uni-versity, Ontario!.

Sherrick, C. E.~1984!. ‘‘Basic and applied research on tactile aids for depeople: Progress and prospects,’’ J. Acoust. Soc. Am.75, 1325–1342.

Sherrick, C. E., and Craig, J. C.~1982!. ‘‘The psychophysics of touch,’’ inTactual Perception, edited by W. Schiff and E. Foulke~Cambridge Uni-versity Press, Cambridge!.

Staller, S. J., Beiter, A. L., and Brimacombe, J. A.~1995!. ‘‘Use of theNucleus 22-channel cochlear implant system with children,’’ Volta RMono. 96, 15–39.

Summers, I. R., and Du, J.~1997!. ‘‘Integration of visual and tactile mo-dalities,’’ Scand. Audiol. Suppl.47, 29–33.

Thornton, A. R., and Raffin, M. J.~1978!. ‘‘Speech discrimination scoresmodelled as a binomial variable,’’ J. Speech Hear. Res.21, 507–518.

Tucker, B. P.~1998!. ‘‘Deaf culture, cochlear implants, and elective disabity,’’ Hastings Cent. Rep.28„4…, 6–14.

Verrillo, R. T., and Gescheider, G. A.~1992!. ‘‘Perception via the sense otouch,’’ in Tactile Aids for the Hearing Impaired, edited by I. R. Summers~Whurr, London!.

Weisenberger, J. M.~1992!. ‘‘Communication of the acoustic environmenvia tactile stimuli,’’ in Tactile Aids for the Hearing Impaired, edited by I.R. Summers~Whurr, London!.

Weisenberger, J. M.~1995!. ‘‘Factors influencing benefit from tactile aids,’Semin. Hear.16, 343–355.

Weisenberger, J. M., Broadstone, S. M., and Saunders, F. A.~1989!.‘‘Evaluation of two multichannel tactile aids for the deaf,’’ J. Acoust. SoAm. 86, 1764–1775.

2979Galvin et al.: Generalization of tactile perceptual skills