Overview and Rationale for Prescriptive Formulas …1 Overview and Rationale for Prescriptive Formulas for Linear and Nonlinear Hearing Aids CATHERINE V. PALMER, GEORGE A. LINDLEY

1Overview and Rationale for Prescriptive

Formulas for Linear and Nonlinear Hearing Aids

CATHERINE V. PALMER, GEORGE A. LINDLEY IV

1

Historical Perspective

As early as 1940, researchers and audiolo-gists were experimenting with selective am-plification. In selective amplification, an at-tempt is made to prescribe a frequencyresponse(s) that is specific to a particular pa-tient’s configuration and degree of hearingloss. Watson and Knudsen (1940) proposed asuprathreshold method for determining re-quired gain and frequency response basedon the patient’s most comfortable loudness(MCL) levels across frequencies. Lybarger(1944) proposed the original -gain rule, in-dicating that approximately half of the hear-ing loss resulted in the appropriate amount ofgain at each frequency in order for conversa-tional speech to be audible and comfortable.Work on developing prescriptive hearing aidfitting methods ended abruptly in 1946 whenthe Harvard Report (Davis et al, 1946) waspublished. This document stated that mostpatients performed best with the same fre-quency response (a flat or 6 dB/octave risingresponse measured in the 2-cc coupler be-tween 300 and 4000 Hz).

Because of the belief that frequency/gaincharacteristics were not important for opti-mizing speech intelligibility, audiologiststurned to a comparative approach for hear-

ing aid selection. Schwartz and Walden(1980) appeared to further support this no-tion when their study revealed that hearingaids with similar electroacoustic characteris-tics differed less than 8% in the majority ofcomparisons using monosyllabic word recog-nition lists. With the same group of subjects,scores on successive tests with the samehearing aids differed as much or more thanthe comparisons between hearing aids.

Until the early 1980s, the primary methodused by audiologists for choosing appropri-ate hearing aids for patients was the com-parative hearing aid evaluation (Humes,1996). In a classic article, Carhart (1946) de-scribed the comparative procedure in whichthe patient is tested with several hearingaids using a variety of speech recognitiontasks. The patient would ultimately be fittedwith the hearing aid(s) that yielded the high-est performance. This method, or a variationthereof, was popular for several decades(Humes, 1996).

In 1983, Walden et al conducted a studythat evaluated key assumptions of the com-parative hearing aid evaluation procedure.These included issues related to sensitivityof testing material, test-retest reliability, andrelationship between relative performancein the sound booth and preference in real-

12

12843.C01.PGS 3/8/02 10:58 AM Page 1

2 STRATEGIES FOR SELECTING AND VERIFYING HEARING AID FITTINGS

world environments. The authors found thatunless hearing aids with very dissimilar fre-quency responses were employed (whichwas unlikely in clinical practice), the as-sumptions of this procedure were not met.

In 1975, Pascoe published an article that re-futed the earlier findings of the Harvard Re-port (Davis et al, 1946). His research, alongwith that of others, indicated that the fre-quency-gain characteristics of hearing aidsare important for optimizing speech intelligi-bility in patients with sensorineural hearingloss. The majority of subjects in the HarvardReport had conductive or mixed hearing loss.At the same time, audiologists were startingto appreciate the difference between couplermeasurements of hearing aid output andreal-ear measurements, including the impactof earmolds on the frequency-gain response(Killion, 1981). All of the data in the HarvardReport were collected through coupler mea-surement. Schwartz and Walden’s (1980) find-ings could be explained by data reported in1978 by Thornton and Raffin that demon-strated that the Northwestern University(NU)-6 word recognition task had very poortest-retest reliability. One could expect thevariability of this measure to obscure real dif-ferences that might exist between hearingaids. The careful work of these individuals aswell as the introduction of custom products,led away from the comparative approach andback to the development of prescriptive hear-ing aid formulas that Lybarger (1944) andWatson and Knudsen (1940) had started 40years earlier.

Introduction to Prescriptive FittingStrategies

The term fitting strategy is analogous to a pre-scription. Simply put, a fitting strategy yieldsamplification settings that are appropriate fora patient based on his or her audiometriccharacteristics. The reason to have a fittingprescription is to generate a target (or targets)of gain as a function of frequency that can beused to select, set, or verify the hearing aid fit-ting in either a coupler or real ear. Fitting strat-egies can be characterized on several dimen-

sions including type of signal processing forwhich they are designed [e.g., linear, wide dy-namic range compression (WDRC)], amountand type of audiometric information required(e.g., audiometric thresholds, dynamic rangemeasures), and underlying theoretical ratio-nale (e.g., loudness equalization, loudnessnormalization) (see rationale section of Table1–1). With a loudness equalization rationale,the goal is to amplify speech sounds such thatthey are perceived as equally loud. With aloudness normalization rationale, the goal isto amplify speech sounds so that normal in-terfrequency loudness relations are main-tained (e.g., vowels are more intense than con-sonants). These rationales are discussed inmore detail in subsequent sections.

Despite differences on one or more of thesedimensions, most fitting strategies have, attheir core, several basic assumptions. Firstand foremost is the recognition that audibilityof speech sounds is critical. This can only beaccomplished via a frequency-gain responsethat has a wide enough bandwidth to amplifycritical high-frequency speech sounds and ac-counts for the degree and configuration of thehearing aid user’s hearing loss. In a series ofinvestigations, Skinner and colleagues dem-onstrated the necessity of a wide frequency re-sponse for maximizing aided speech recogni-tion ability (Skinner, 1980; Skinner and Miller,1983).

Restoration of audibility, however, takesplace within the constraints of the patient’sdynamic range and sound quality percep-tions. Dynamic range is the area between justaudible sounds (threshold) and sounds thatare judged to be uncomfortable. There isa need to recognize the functional limitationsimposed by the dynamic range of the listenerwhen prescribing amplification character-istics. See Chapter 2 for a comprehensiveoverview on measuring loudness discomfortlevel (LDL) and for selecting/verifying theappropriate output of the hearing aid basedon LDL data. As a general rule, the amplifiedsignal should not exceed the patient’s LDL.With regard to sound quality, it is evident thatcertain frequency-gain responses, althoughappropriate for restoring audibility, may re-

12843.C01.PGS 3/8/02 10:58 AM Page 2

CHAPTER 1 ■ PRESCRIPTIVE FORMULAS FOR LINEAR AND NONLINEAR HEARING AIDS 3Ta

ble

1–1

.C

omp

aris

on a

mon

g th

e W

ide

Dyn

amic

Ran

ge C

omp

ress

ion

(WD

RC

) Fit

tin

g S

trat

egie

s D

iscu

ssed

in T

his

Ch

apte

r

NA

L-N

L1

(Nat

iona

l D

SL[i

/o] (

Des

ired

IH

AFF

Fitt

ing

Aco

usti

c L

abor

ator

ies

Sens

atio

n L

evel

(I

ndep

ende

nt H

eari

ng

Stra

tegy

Non

line

ar V

ersi

on 1

)[I

nput

/Out

put]

)FI

G6

Aid

Fit

ting

For

um)

LE

Thr

esho

ld d

ata*

Cou

ple

r ta

rget

s, r

eal-

ear

aid

ed ta

rget

s, r

eal-

ear

gain

targ

ets,

rea

l-ea

rin

sert

ion

gain

, aid

edth

resh

old

s, S

pee

ch-O

-G

ram

, HA

par

amet

ers

(e.g

., C

T, C

F, C

R)

Use

r ch

oose

s be

twee

n 1

and

4 ch

anne

ls

RE

CD

, RE

UG

, HA

styl

e, C

T,ea

rmol

d o

pti

ons,

ou

tpu

tlim

itin

g ch

arac

teri

stic

sYe

s

Yes

(pro

vid

es c

omm

ents

rela

ted

to c

hoic

es)

Yes,

can

ent

er b

irth

dat

e

Yes

Nat

iona

l Aco

ust

icL

abor

ator

ies

(Au

stra

lia)

ww

w.n

al.g

ov.a

u

Rat

iona

le:

Au

dio

met

ric

info

rmat

ion

that

can

be

use

d

Fitt

ing

info

rmat

ion

pro

vid

ed

Cha

nnel

s

Ind

ivid

ualiz

ed H

Afi

ttin

gd

ata

that

can

be

ente

red

:

Adj

ustm

ent f

or b

inau

ral

fittin

g?A

dju

stm

ent f

or v

ent a

ndtu

bing

cho

ices

?A

ge-r

elat

ed a

dju

stm

ents

for

RE

CD

?A

dju

stm

ent f

orco

ndu

ctiv

e lo

ss?

Con

tact

info

LE

or

LN

**T

hres

hold

dat

a*, L

DL

s,lo

ud

ness

con

tou

r d

ata

Cou

ple

r ta

rget

s, r

eal-

ear

aid

ed ta

rget

s, r

eal-

ear

gain

targ

ets,

aid

edth

resh

old

s, S

PL

-O-g

ram

,H

Ap

aram

eter

s (e

.g.,

CT,

CF,

CR

)

Pro

vid

es in

form

atio

n fo

r 9

chan

nels

RE

CD

, RE

UG

, RE

DD

, HA

styl

e, e

arm

old

op

tion

s,C

T, c

omp

ress

ion

typ

eN

o

Yes

Yes,

mu

st e

nter

bir

th d

ate

No,

bu

t tar

gets

acc

ura

te if

the

ind

ivid

ual

’s L

DL

sar

e en

tere

dN

atio

nal C

ente

r fo

rA

ud

iolo

gy (C

anad

a)w

ww

.uw

o.ca

/nca

LN

Thr

esho

ld d

ata

Cou

ple

r ta

rget

s, r

eal-

ear

inse

rtio

n ga

in ta

rget

s H

Ap

aram

eter

s (e

.g.,

CT,

CF,

CR

)

Ass

um

es 2

cha

nnel

s

Non

e

No

No

No

No

Ety

mot

ic R

esea

rch

ww

w.e

tym

otic

.com

LN

Lou

dne

ss c

onto

urs

or

thre

shol

d d

ata

wit

hco

ntou

r d

ata

pre

dic

ted

i/

o cu

rves

, HA

par

amet

ers

(e.g

., C

T, C

F, C

R)

Ass

um

es 2

cha

nnel

s, b

ut

use

r ca

n ge

t inf

o fr

omev

ery

cont

our

freq

uen

cy(u

p to

6 fr

equ

enci

es)

Non

e

Yes

No

No

No,

bu

t tar

gets

acc

ura

te if

lou

dne

ss c

onto

ur

dat

aen

tere

dH

eari

ng A

id R

esea

rch

Lab

ww

w.a

usp

.m

emp

his.

edu

/ha

rl/

* O

btai

ned

und

er h

ead

pho

nes,

inse

rt e

arp

hone

s, o

r in

sou

nd fi

eld

.

** S

ee te

xt.

LD

L, l

oud

ness

dis

com

fort

leve

l; R

EC

D, r

eal-

ear-

to-c

oup

ler

dif

fere

nce;

RE

UG

, rea

l-ea

r-u

naid

ed-g

ain;

RE

DD

, rea

l-ea

r-to

-dia

l-d

iffe

renc

e; C

T, c

omp

ress

ion

thre

shol

d; C

R, c

omp

ress

ion

rati

o; L

E,

lou

dne

ss e

qual

izat

ion;

LN

, lou

dne

ss n

orm

aliz

atio

n; C

F, c

ross

over

freq

uen

cy; i

/o,

inp

ut/

outp

ut;

HA

, hea

ring

aid

12843.C01.PGS 3/8/02 10:58 AM Page 3


sult in sound quality perceptions that are un-acceptable. As demonstrated by Skinner et al(1982a), for example, this is a possibility whena sharply rising frequency-gain response isprescribed. Work by Hogan and Turner (1998)and Ching et al (1998) indicates that high-frequency amplification actually may be detri-mental in terms of speech intelligibility forsome users depending on the degree and siteof lesion of the hearing loss. As such, soundquality and distortion must be consideredwhen determining appropriate fitting targets.The prescription generated must reflect a bal-ance among audibility, loudness, and soundquality (Byrne, 1992). Within the frameworkoutlined above, the best method for determin-ing appropriate amplification characteristicsremains unclear. Based on the large number offitting strategies available to the audiologisttoday, it is evident that a consensus has notbeen reached, and empirical support for anyparticular approach(es) is lacking.

Practical Considerations

In choosing a fitting strategy, the audiologistmay want to consider the following:

1. How much audiometric and related in-formation will be employed within theprescriptive procedure in generating thefitting targets?

2. Does the prescriptive procedure use age-specific corrections if individual data arenot available?

3. What types of transducers can be usedto obtain the audiometric data?

4. What are the resulting targets that areprovided for selection, presetting, andverification?

Table 1–1 compares four prescriptive strate-gies with regard to the practical considera-tions described below.

Audiometric Information That Can Be Used

The audiologist may want to work with a fit-ting strategy that requires minimum audio-metric data (i.e., thresholds only; see Table1–1, audiometric information that can beused). This implies that the fitting program

will estimate LDL, MCL, and/or loudnesscontour data. The research available regard-ing the relationship between threshold andMCL and LDL indicates that the relationshipis not strong (Kamm et al, 1978; Cox andBisset, 1982), and that at least 25% of patientswill deviate substantially from average interms of LDLs (Pascoe, 1990). Using a thresh-old-based formula implies that the audiolo-gist is making a choice as to where to spendtime in the fitting process and is using flexi-ble (programmable) technology that can bemanipulated during the fitting process. Timecan be spent in measuring MCL and LDLand including this in a fitting strategy priorto hearing aid fitting, or time can be spentduring the fitting to fine-tune the hearingaids to account for any patient differencesthat were not considered in the audiometrictesting. If the latter is chosen, the hearingaids must be flexible enough to change basedon verification results, and the patient mustbe able to provide subjective feedback thatcan be used in determining appropriate ma-nipulations of programmable parameters.

Individualized Hearing Aid and EarmoldFitting Data That Can Be Entered

The audiologist working with infants andchildren may wish to use a fitting protocolthat accounts for as much individual data aspossible [e.g., real-ear-to-coupler difference(RECD); see Chapter 3 for a full descriptionof real-ear measures] because a young childcannot provide subjective feedback for fine-tuning purposes. Age-specific correctionsmay be useful with a pediatric population aswell. In other words, if the RECD cannot bemeasured, does the fitting protocol provideestimates of this measure based on the age ofthe child who will be receiving the hearingaids? The audiologist may want to workwith a fitting strategy that allows audiomet-ric data to come from a variety of transducers(e.g., sound field, insert earphones, head-phones, etc.). This is important when work-ing with a pediatric population, as we cannotalways assess children in a manner similar tothat for adults (Table 1–1, age-related adjust-ments for RECD)

12843.C01.PGS 3/8/02 10:58 AM Page 4

CHAPTER 1 ■ PRESCRIPTIVE FORMULAS FOR LINEAR AND NONLINEAR HEARING AIDS 5

Entry of data related to the style of the ear-molds and/or circuitry type of the hearingaids that will be employed can further indi-vidualize the fitting. The impact of style andventing characteristics can be incorporatedwhen determining appropriate coupler set-tings that should yield real-ear measures thatwill closely approximate the prescriptive tar-get(s) with minimal adjustment. Compres-sion characteristics available on the particu-lar hearing aids [e.g., compression threshold(CT), compression ratio (CR)] can be consid-ered a priori in deriving appropriate fittingtargets. If this information cannot be entered,compromises may need to be made whentrying to meet the targets. Table 1–1 (individ-ualized HA fitting data that can be entered)describes the user’s ability to use individualdata in each fitting formula.

Fitting Information Provided

The audiologist may select a fitting strategybased on what verification methods are pro-vided. If the audiologist prefers to verify thehearing aid fitting with real-ear measure-ments, a protocol that provides real-ear in-sertion gain (REIG), real-ear aided response(REAR), or real-ear aided gain (REAG) tar-gets may be preferred. See Chapter 3 for de-tails related to real-ear measures. The audi-ologist may want coupler gain or outputtargets to select conventional hearing aidsbased on matrix data or to preset program-mable hearing aids prior to real-ear verifica-tion (Table 1–1, fitting information provided).

Channels

Prescriptive strategies can differ on the num-ber of channels for which compression char-acteristics are prescribed. With multichannelhearing aids, the gain, CR, and CT can bemanipulated individually in each channel.As such, with increasing circuitry sophistica-tion, greater prescriptive data are needed(Table 1–1, channels).

Adjustments for Binaural Fitting

When prescribing targets for a binaural ver-sus a monaural fitting, the impact of binaural

summation should be considered. With abinaural fitting, the audiologist can expect a3- to 6-dB reduction in the amount of gainnecessary to achieve a given loudness whencompared with a monaural fitting. Withsome prescriptive strategies, this reduction isautomatically incorporated when a binauralfitting is chosen. With others, the audiologistmust decide whether or not to reduce theamount of gain and to what degree (Table1–1, adjustment for binaural fitting?).

Adjustments for Conductive Loss

With a purely conductive hearing loss, there isno reduction in the patient’s dynamic range.Therefore, a linear processing scheme may bemore appropriate as a constant amount ofgain is needed as a function of increasinginput level. With sensorineural hearing loss,the greatest amount of gain is required for softinput levels and progressively less gain isneeded with increasing input level. When fit-ting a conductive or mixed hearing loss, use ofa prescriptive strategy that accounts for anyconductive component is warranted (Table1–1, adjustment for conductive loss?).

Theoretical Considerations

In reviewing the underlying theory andmethods for implementing various prescrip-tive strategies, it is evident that these pre-scriptive strategies differ on primarily oneor two dimensions. The first dimension re-lates to how the frequency bands within anamplified signal should relate with regard tooverall loudness (Byrne, 1996). A second di-mension relates to how the fitting strategy isincorporated within a fitting protocol (Lind-ley et al, 2001). These issues are discussed inmore detail in the following sections.

Loudness Normalization (LN) versusLoudness Equalization (LE)

The underlying theory behind most prescrip-tive strategies can be described as a loudnessnormalization (LN) rationale or a loudnessequalization (LE) rationale. With an LN ra-tionale, the overall goal is to restore normal

12843.C01.PGS 3/8/02 10:58 AM Page 5


loudness perception for the patient wearinghearing aids. For example, the speech pre-sentation levels judged to be “soft,” “com-fortable,” and “loud” by normal hearingindividuals also should be judged “soft,”“comfortable,” and “loud” by the hearing aidwearer. In addition, the normal relative loud-ness across frequency bands in a complexsignal should be maintained. With a speechsignal, for example, this means that at a givenpresentation level, the lower frequencies willbe perceived at a greater loudness than thehigher frequencies (i.e., vowels have moreacoustic power than consonants). The LN ra-tionale is based on the assumption that resto-ration of normal loudness perception willlead to a more successful fitting and greateracceptance by the hearing aid user. Rela-tively greater weight is placed on loudnessand sound quality as opposed to maximizingspeech intelligibility.

With an LE rationale, the goal is to equal-ize loudness across frequency bands. Thisoccurs while normalizing overall loudness.Thus, the patient should perceive the overallspeech levels in a manner similar to that ofindividuals with normal hearing. It is the re-lation between frequency bands within thesignal that differ. For a conversational inputlevel [i.e., 65 dB sound pressure level (SPL)],for example, the various frequency bandswould be perceived as being equally loud.The development of the National AcousticLaboratories’ (NAL) fitting procedure hasstimulated considerable research, the resultsof which suggest that hearing aid settingsthat equalize perceived loudness across fre-quency bands are preferred by many pa-tients with hearing loss (Byrne, 1986). Oneassumption of this rationale is that speechrecognition will be higher than that obtainedwith an LN rationale.

Incorporation of a Fitting Strategywithin the Fitting Protocol

Regardless of what fitting strategy is em-ployed in deriving an initial prescription, it isimportant to remember that the fitting strat-egy is only part of the whole fitting process.

Employment of a fitting strategy results in aprescription for how the hearing aids shouldfunction when the output or gain of the hear-ing aids are measured in a coupler or in theuser’s ear canal. What data are employed inderiving the initial targets and how far oneinitially deviates from this prescription whenfitting hearing aids depends on what type offitting protocol is employed.

At one extreme, an audiologist-driven(AD) protocol may be employed (Lindley etal, 2001). With an AD protocol, only thresh-old data are used with the fitting strategy inderiving a prescription. On the day of the fit-ting, the audiologist attempts to meet thetarget(s) as closely as possible. Further ad-justments are not made to the hearing aidsbased on initial subjective perceptions pro-vided by the hearing aid user. Rather, the as-sumption is that the prescription employedrepresents the best compromise betweenspeech recognition and sound quality. Thehearing aid user is provided with the oppor-tunity to adapt to these hearing aid settingsover a given period of time before anychanges are considered. With this protocol,speech recognition is given higher prioritythan initial perceptions of sound quality orloudness.

In contrast to the AD protocol, a patient-driven (PD) protocol requires significantinput regarding initial perceptions of loud-ness and sound quality from the patientbeing fitted with hearing aids (Lindley et al,2001). The patient provides input at twostages of the fitting process. In stage 1, au-diometric information such as thresholds,loudness contours, and/or LDLs are mea-sured to determine the appropriate hearingaid settings. In stage 2, loudness judgmentsand formal or informal sound quality rat-ings are obtained on the day of the fittingwith the patient wearing the hearing aids.This information is employed to verify thatthe original fitting goals have been met. Ifloudness normalization is the goal, the judg-ments obtained from the patient can be com-pared with normative data to ensure thatnormal loudness perception has been re-stored. With a PD protocol, initial percep-

12843.C01.PGS 3/8/02 10:58 AM Page 6


tions of sound quality and/or loudness aregiven higher priority than maximization ofspeech recognition.

In clinical settings, it is likely that a combi-nation of the above protocols is employed.That is, adjustments are often made on theday of the fitting to increase sound comfortyet maintain adequate, if not ideal, audibil-ity levels. The use of adaptation managerswithin some manufacturers’ fitting softwarefacilitates a gradual introduction to amplifi-cation, especially in first-time users. Hearingaid settings also can be adjusted based onthe amount of hearing aid experience of theuser. Over time, the amount of amplificationprescribed, and hence audibility, is increasedin an effort to improve speech recognitionability while providing acceptable loudnessand sound quality. In this type of protocol,the targets provided by a given fitting strat-egy might serve as an ultimate goal, ratherthan an initial setting.

Device-Independent versus ProprietaryFitting Strategies

Device-independent fitting strategies implythat the prescriptive formula is not hearingaid dependent. The audiologist enters the re-quired patient data and is presented with

coupler or real-ear target(s) that can be ap-plied to any hearing aid. The target(s) mayinclude gain as a function of frequency, gainas a function of input level (for nonlineartechnology), desired CT(s), and CR(s). TheCT is the input level at which the signal isprocessed in a nonlinear, as opposed to lin-ear, manner. In Figure 1–1 this occurs at 40-dB SPL for the WDRC example. The CR rep-resents the degree of compression that isapplied to the incoming signal and is calcu-lated by determining how much of an in-crease in the amplified signal occurs whenan increase in the input signal occurs. In theWDRC example from Figure 1–1, as theinput level is increased from 40- to 70-dBSPL, the amplified signal increases 10 dB, re-sulting in a CR of 3:1 [i.e., change in input(30) divided by change in output (10)].

Proprietary fitting strategies are related toa particular circuit/technology. Individualpatient data are entered (threshold, LDL,etc.), and a proprietary fitting algorithm pro-grams all of the signal processing features ofthe hearing aids. This may include compres-sion characteristics (threshold, ratio, attackand release times) in each channel as well asfrequency-gain response and output as afunction of changes in the overall inputlevel. A variety of manufacturers have de-

Figure 1–1. Input/output graphs for linear and wide dynamic range compression(WDRC) signal processing. SPL, sound pressure level.

12843.C01.PGS 3/8/02 10:58 AM Page 7


rived their own fitting algorithms as theymoved toward digital signal processing(DSP). The rationale is that there is virtuallyinfinite flexibility in DSP, and therefore themanufacturer must constrain some of theprogrammability. At a minimum, the audiol-ogist needed a starting point with thesetechnologies and did not desire a truly openplatform technology where all aspects of sig-nal processing could be manipulated by theaudiologist. Rather a closed DSP platform isemployed with the audiologist able to haveprogrammable control only over specific sig-nal processing characteristics. The advanceduse of WDRC in achieving audibility through-out the dynamic range and reducing noise in-terference also pointed toward proprietary al-gorithms that set the compression timeconstants. None of the device-independentfitting strategies provides guidelines as tohow to control the compression time con-stants across the various channels of the hear-ing aid.

A few manufacturers have provided soft-ware that represents a compromise betweenproprietary fitting algorithms and generallyaccepted independent fitting strategies. Forexample, the software for the Siemens Signiaallows the audiologist to choose between adesired sensation level (DSL) input-output(i/o)-based fitting (emphasis on audibility)and an NAL nonlinear version 1 (NL1)-based fitting (emphasis on speech intelligi-bility within the constraints of loudness,sound quality, and degree of hearing loss)(Cornelisse et al, 1995; Byrne et al, 2001).Within the independent strategies, propri-etary programming of the instrument is car-ried out related to compression characteris-tics (e.g., attack and release times, CT in eachchannel, etc.). The software for the Quali-tone programming screen allows choicesamong a variety of the well-known fitting al-gorithms [FIG6, visual input-output locatoralgorithm (VIOLA), etc.] so as to leave thefinal rationale to the audiologist while set-ting some parameters in a proprietary man-ner (Gitles and Niquette, 1995; Cox andFlamme, 1998). Many manufacturers alsoprovide the audiologist with the option of

choosing among two or more levels of pro-gramming flexibility (i.e., number of para-meters that can be manipulated) when ad-justing the hearing aids. Audiologists canwork at the programming level at whichthey are most comfortable.

Regardless of which type of prescriptivestrategy (i.e., proprietary versus indepen-dent) and the degree to which the audiolo-gist can control various signal processingcharacteristics, verification that fitting ob-jectives have been met is critical. Real-earmeasures related to the patient’s dynamicrange, for example, are a powerful tool thatcan be used to document that audibilityhas been restored and the patient’s LDLhas not been exceeded. The methodologyused to reach the final amplification set-tings is secondary to achieving the desiredresults.

Linear and Nonlinear DeviceIndependent Fitting Strategies

A linear hearing aid fitting is two-dimen-sional. Gain varies as a function of frequency.Gain, however, does not vary as a function ofinput level; therefore, a single frequency-gain target is produced by the fitting for-mula. The gain will remain the same regard-less of input level until the hearing aids reachsaturation. The input/output function forthis type of signal processing will reveal astraight line at 45 degrees where a change ininput results in an identical change in output(a one-to-one ratio; see Figure 1–1). The audi-ologist is aware that regardless of meetingprescriptive targets, the user of linear ampli-fication can choose to use less or more gainas a function of the user-manipulated vol-ume control wheel.

A hearing aid fitting that includes com-pression signal processing in the dynamicrange (e.g., WDRC), on the other hand, isthree-dimensional. Gain varies as a functionof frequency and input level. Prescriptive fit-ting formulas for this type of signal process-ing will produce at least three targets, typi-cally for soft (e.g., 45-dB SPL), moderate(e.g., 65-dB SPL), and loud (e.g., 85-dB SPL)

12843.C01.PGS 3/8/02 10:58 AM Page 8


input levels. The gain generated for softinput levels would be greater than the gaingenerated for moderate or loud inputs lev-els. Many of these instruments are dis-pensed without a volume control becausethe one consistent advantage of WDRC isthe automatic nature of the instrument (alle-viating the user of the need to increase thegain for soft sounds and decrease the gainfor loud sounds). The input/output functionfor this type of signal processing will reveala linear response up to the compressionthreshold (kneepoint) and then changes ininput will not produce identical changes inoutput. From Figure 1–1, it is evident thatmore of the aided signal falls within the pa-tient’s dynamic range (shaded region repre-sents the auditory area between thresholdand LDL) with WDRC processing as op-posed to linear processing. Alternatively, ex-pansion (providing a decreasing amount ofgain with decreasing input level) can be em-ployed below the compression threshold inan attempt to reduce the annoyance of low-level ambient noises.

Examples of Fitting Strategies Designedfor Linear Signal Processing

Of the many prescriptive formulas proposedfor linear amplification over the years, only afew are in general use, and those will be re-viewed in the following section. The formu-las for these fitting strategies are listed inTable 1–2 and include National Acoustic Lab-oratories-Revised (NAL-R; Byrne and Dillon,

1986), prescription of gain and output(POGO; McCandless and Lyregaard, 1983),Libby one-third gain (Libby, 1986), and Berger(Berger et al, 1989). The desired sensationlevel (DSL, Seewald, 1992, 1994, Seewald et al1995) also provides a fitting strategy for linearamplification. The DSL procedure is not asimple multiplication based on magnitude ofhearing loss. Thresholds in dB hearing level(HL) are converted to dB SPL using the mea-sured or predicted RECD, and a speech spec-trum (adult or child) chosen by the user isused to create the final target levels for theamplified speech spectrum.

The Berger et al (1989) procedure is basedon the assumption that amplification shouldincrease the input levels toward averageconversational speech levels (65-dB SPL)and that amplification at 250 to 500 Hzshould be reduced slightly because of theirpotential detrimental masking effect onspeech recognition. The formula promotedby Libby (1986) is based on the finding thatpatients with mild hearing loss use very lit-tle gain in real-life listening situations. Heconcluded that a prescription should reflectthe amount of gain actually used by the sub-jects in his study. The general rule of theNAL-R (Byrne and Dillon, 1986) is to pro-vide sufficient gain at each frequency to am-plify average conversational speech to thepatient’s MCL. The NAL-R approach at-tempts to control for excessive gain in caseswith steeply sloping hearing loss by provid-ing slope gain modifications by multiplyingthe HL at each frequency by 0.31. The POGO

Table 1–2. Multipliers Applied to Thresholds to Calculate the Gain Needed to AllowConversational Level Speech to be Audible and Comfortable in Four Threshold-Based FittingProcedures Designed for Linear Amplification

250 Hz 500 Hz 1000 Hz 2000 Hz 3000 Hz 4000 Hz 6000 Hz

NAL-R* 0.31 (�17) 0.31 (�8) 0.31 (+1) 0.31 (�1) 0.31 (�2) 0.31 (�2) 0.31 (�2)POGO 0.5 (�10) 0.5 (�5) 0.5 0.5 0.5 0.5 0.5Libby 1/3 0.33 (�5) 0.33 (�3) 0.33 0.33 0.33 0.33 0.33Berger — 0.30** 0.63 0.67 0.59 0.53 0.50

* Plus 0.05 of hearing level at 500 + 1000 + 2000 Hz (addition for overall gain).

**0.50 for a 50-dB hearing loss.

NAL-R, National Acoustic Laboratories–Revised; POGO, prescription of gain and output.

12843.C01.PGS 3/8/02 10:58 AM Page 9


procedure (McCandless and Lyregaard, 1983)is based on the assumption that frequency-gain response and saturation sound pres-sure level (SSPL) limiting are essential char-acteristics in a basic prescriptive fitting.Similar to the other strategies, POGO’s goalis to create audible, comfortable conversa-tion level inputs. It is worth noting that max-imum output specification is essential to theBerger, Libby, and POGO formulas.

Figure 1–2 illustrates the coupler gain tar-gets generated for a mild (25 dB HL at 250Hz) to severe (90 dB HL at 6000 Hz) slopinghearing loss from the DSL, NAL-R, POGO,Berger, and Libby fitting strategies. It is evi-dent that the POGO and DSL fitting algo-rithms prescribe more gain in the middle tohigh frequencies. Whether these amplifica-tion levels can be obtained without feedbackand with user satisfaction is a worthwhileconcern, but the goal is to provide audiblesignals across a wide frequency range. Sev-eral of the fitting programs were modified inlater years to provide slight variations formore profound hearing losses (e.g., NAL-RP,Byrne et al, 1990; POGO II, Schwartz et al,1988; Libby two-thirds gain rule, Libby, 1986).Also, DSL and NAL prescriptive formulasnow provide advanced fitting programs fornonlinear technology (NAL-NL1, Byrne et

al, 2001; and DSL [i/o], Cornelisse et al, 1995).These updated versions are discussed in thenext section.

As mentioned earlier, linear prescriptivemethods produce a single gain target (re-lated to the coupler or real ear) because ofthe nature of linear amplification. Gain doesnot vary as a function of input level, so a sin-gle target tells the whole story. The gain isprescribed to optimize conversation level in-puts, acknowledging that soft sounds will beinaudible and that the user will operate atthe limits of the hearing aids in many situa-tions (above conversation level inputs). Ifaverage conversational speech is at a com-fortable loudness level, the linear fitting istypically successful. The user is providedwith a volume control that would have to bemanipulated very rapidly to keep the vari-ous input levels one expects in conversationand in the environment to be maintained inthe audible range.

Considering that the user has access to avolume control in linear hearing aid fittings,differences in frequency-gain responses be-tween prescriptive formulas would be mini-mized (Humes, 1986, 1988; Sullivan et al,1988). Byrne and Cotton (1987) have argued,however, that even when allowing for pa-tient volume control adjustments, speech

Figure 1–2. 2-cc coupler gain for five linear fitting strategies. DSL, desired sensation level;NAL-R, National Acoustic Laboratories-Revised; POGO, prescription of gain output; SPL,sound pressure level.

12843.C01.PGS 3/8/02 10:58 AM Page 10


recognition is significantly better with someprescribed target criteria than with others.That is, some targets, even after beingequated for loudness, are expected to yieldhigher speech recognition.

Suprathreshold Fitting Strategies

In the past, fitting strategies have been di-vided into threshold-based and suprathresh-old-based. Watson and Knudsen (1940) usedMCL measures in their fitting strategy andwere the first group to use suprathresholdmeasures in a prescriptive formula. Cur-rently, this is an arbitrary division becauseseveral strategies may be used either way(e.g., VIOLA, DSL [i/o]) (Cornelisse et al,1995; Cox and Flamme, 1998). These ap-proaches accept suprathreshold data (i.e.,loudness contour, LDL), but also will predictthese measures if only hearing thresholds areprovided. Table 1–3 provides a descriptionof several suprathreshold fitting strategies.The CID (Central Institute for the Deaf, Skin-ner et al, 1982b; Skinner, 1988) and Shapiro(Shapiro, 1976, 1980) methods employ MCLmeasures. The MSUv3 (Memphis State Uni-

versity; Cox, 1988) and Bragg (Bragg, 1977)methods bisect the dynamic range, thereforerequiring threshold and LDL data. The Levittmethod (Levitt et al, 1987) bases recommen-dations on LDL measurements. Some re-searchers have reported that some patientsdo not produce reliable MCLs (Stephens etal, 1977; Berger and Soltisz, 1981; Skinnerand Miller, 1983), which puts methods basedon MCL measurement in question. Palmerand Lindley (1998) found that full loudnesscontours (threshold to LDL) were reliable ina group of subjects. Based on a review of re-cent publications and presentations, none ofthese strategies appears to be in general use,but knowledge of the premise behind each(see Table 1–3) assists the student of prescrip-tive formulas in understanding the most re-cent (nonlinear) approaches.

Examples of Fitting Strategies Designedfor Nonlinear Signal Processing

In this section, four fitting strategies de-signed for hearing aids incorporating non-linear processing are discussed: NAL-NL1,DSL [i/o], FIG6, and VIOLA. These strate-

Table 1–3. Summary of Suprathreshold-Based Prescriptive Methods

Method Objective References

CID Amplify speech to MCL (gain reduced at Skinner et al, 1982b250 Hz by a variable amount); Skinner (1988) Skinner, 1988recommended amplifying speech to halfwaybetween threshold and MCL for 250 Hz and 6000 Hz; current recommendation is amplification at 90% of the range from threshold to MCL at 1000–2000 Hz and 80% of this range at 2000–4000 Hz

Shapiro 60 dB SPL pure tones amplified Shapiro, 1976, 1980(gain reduced by 15 dB and 10 dB at 250 and 500 Hz, respectively)

MSUv3 Amplify speech to a level midway between Cox, 1988threshold and LDL for 250–6000 Hz

Bragg Amplify speech halfway between threshold Bragg, 1977and LDL at 1000 Hz and above andone-third of this range at 250 and 500 Hz

Levitt Amplify speech to a level 10 dB below LDL Levitt et al, 1987between 1000 and 6000 Hz, and 22and 16 dB below LDL at 250 and 500 Hz, respectively

CID, Central Institute for the Deaf; LDL, loudness discomfort level; MCL, most comfortable loudness level;SPL, sound pressure level.

12843.C01.PGS 3/8/02 10:58 AM Page 11


gies were chosen as they are device indepen-dent, are currently available in many hearingaid fitting software packages, and presentdifferent theoretical rationales and imple-mentation approaches. For each, a descrip-tion of the underlying theory and implemen-tation is provided. Additional informationregarding procedural aspects is provided inTable 1–1.

NAL-NL1

The National Acoustic Laboratories’ nonlin-ear fitting procedure, version 1 (NAL-NL1)is an extension of the NAL-R fitting strategyfor linear amplification (Byrne et al, 2001).NAL-NL1 was designed for nonlinear signalprocessing and provides different prescrip-tive targets as a function of input level. WithNAL-NL1, the goal is to maximize speechrecognition for a given speech input level(dB SPL) within the constraint that the hear-ing aid wearer should perceive the overalllevel of speech no louder than an individualwith normal hearing. Speech judged to be“soft,” “comfortable,” and “loud” by indi-viduals with normal hearing should be per-ceived similarly by the patient wearing hear-ing aids. At a given input level, however, therelative loudness across frequency regions isprescribed with the goal of maximizing in-telligibility. At least for conversational inputlevels (i.e., 60- to 70-dB SPL) this typicallyleads to prescriptions similar to NAL-R andsignificantly different from prescriptionswith an LN rationale. As such, this fittingstrategy would best be described as havingan LE rationale.

In developing the formula employedby this strategy, a modified version of theSpeech Intelligibility Index (SII) was used inconjunction with a loudness model to meetthe goals of maximizing intelligibility whilemaintaining normal overall loudness percep-tion (ANSI, 1997; Byrne et al, 2001). A uniqueaspect of NAL-NL1 is that the impact of highsensation levels and degree of hearing loss,with regard to speech recognition, is takeninto account. Functionally, this results in lessprescribed gain in regions of severe hearingloss than would be expected if audibility

alone were considered. The NAL-NL1 alsoallows the user to enter the birth date of thepatient to use age-appropriate correction fac-tors (e.g., for RECD) or to enter these data di-rectly. This fitting version is greatly en-hanced compared with NAL-R and could beused with a pediatric population.

The audiologist enters the audiometric andhearing aid data (e.g., style, number of chan-nels, etc.) into the fitting software. Additionalindividual-specific data such as RECD orreal-ear unaided response (REUR) can be en-tered as well. Figure 1–3 provides an exampleof the printout provided by the NAL-NL1software. The audiologist can choose to viewthe target data in a variety of forms including2-cc coupler gain and REIG. Ideal crossoverfrequency(ies), CT(s), and CR(s) are providedin tabular format to facilitate adjustment ofthe hearing aids to meet the desired targetsand to aid in choosing appropriate hearingaids. These values can be overridden if de-sired, and the targets are automatically recal-culated accordingly.

DSL [i/o]

The goal of the DSL [i/o] hearing aid fittingprocedure is to prescribe amplification char-acteristics such that the entire range of acous-tic signals available to a patient with normalhearing is placed within the dynamic rangeof the patient with hearing loss (Cornelisse etal, 1995). The underlying rationale, with re-gard to relative loudness across frequencybands, varies according to the parameterschosen. For example, when the variable com-pression (nonlinear) ratio mode (CR changesas a function of input level) is chosen, the tar-gets generated are based on an LN rationale.When linear compression ratio mode is cho-sen (CR remains constant above the CT), thetargets generated are based on an LE ratio-nale (Seewald et al, 1997). As outlined earlier,with the LE rationale, the goal is to provideamplification such that relative loudnessacross frequency bands is equalized. Thelong-term average speech spectrum (LTASS)should fall along the individual’s MCL levelacross frequency. Thresholds and the upperlimit of comfort (either measured or pre-

12843.C01.PGS 3/8/02 10:58 AM Page 12


dicted) across frequency are used to definethe dynamic range.

The extent to which the DSL [i/o] goal canbe met depends on the compression thresh-old. Input levels that fall below the compres-

sion threshold will not be audible for the pa-tient with hearing loss. Given that the lowestcommercially available compression thresh-old is approximately 30-dB SPL, it is appar-ent that the goal of providing all acoustic

Figure 1–3. Example of prescriptive output provided by National Acoustic Laboratoriesnonlinear version 1 for a 2-channel hearing aid (NAL-NL1).

12843.C01.PGS 3/8/02 10:58 AM Page 13


signals available to normal hearing patientsto patients with hearing loss is not likely.The use of nonlinear technology, however, inconjunction with an appropriate fitting strat-egy such as DSL [i/o], allows one to func-tionally approach this goal.

The DSL [i/o] allows for as much or as littleindividual data (e.g., LDL, RECD) to be en-tered as desired by the audiologist and pro-vides age-appropriate corrections for RECDand LDL. These elements and the choice ofmultiple transducers (Table 1–1) make thisfitting strategy very popular for use with in-fants and young children, but also appropri-ate for an adult population. Target data canbe viewed in a variety of formats.

In Figure 1–4, for example, coupler verifi-cation data for a variety of input levels areprovided. The software also lists recom-mended CR(s) and SSPL90 as a function offrequency. Verification data for real-ear mea-sures also can be provided. Alternatively, theaudiologist (and patient) can view the targetdata in the form of an SPL-O-Gram (Fig.1–5). The patient’s threshold and LDL data,as well as aided output targets, actual aidedoutput values, and/or unaided output val-ues, can be viewed on one graph. This can bea useful counseling tool, as the patient cansee which sounds are inaudible when un-aided, and the improvement in audibility af-forded by the hearing aids. Several addi-tional output formats are available.

FIG6

The FIG6 fitting strategy employs a loud-ness normalization rationale (Gitles and Ni-quette, 1995). Only threshold data are em-ployed in generating fitting targets. FIG6provides targets for low (40-dB SPL), middle(65-dB SPL), and high (90-dB SPL) input lev-els. For low input levels, gain is prescribedwith the goal of providing aided thresholdsof 20 dB HL when the hearing loss is mild-to-moderate in severity. For greater degreesof hearing loss, aided thresholds increase asa function of increasing severity. At middleto high input levels, the gain required fornormal loudness perception assuming 65-and 95-dB SPL input levels is prescribed.

The amount of gain required is based on av-erage data obtained from individuals withhearing loss.

FIG6 is a good example of a fitting proto-col that assumes the audiologist is workingwith an adult who can provide feedback andwith flexible technology that can be adjustedto account for individual differences thatwere not accounted for in any premeasure-ments (e.g., RECD, LDL). As seen in Figure1–6, REIG and coupler targets are providedas well as CRs in a low- and high-frequencychannel.

VIOLA

VIOLA is part of the Independent HearingAid Fitting Forum (IHAFF) protocol for theselection, fitting, and verification of hearingaids (Valente and Van Vliet, 1997; Cox andFlamme, 1998). VIOLA employs a loudnessnormalization rationale, and the IHAFF pro-tocol is best described as patient-driven.Loudness rating data for warble tones ob-tained using the Contour Test of LoudnessPerception are used in deriving the ultimatefitting targets (Cox et al, 1997). The currentversion of the fitting protocol will estimatethe loudness contours if only threshold dataare available.

The patient rates the loudness of warbletones using a seven-point scale ranging from“very soft” to “uncomfortably loud.” An as-cending approach is employed, and severaltrial runs are obtained at each frequency(typically 500 and 3000 Hz at a minimum).VIOLA uses this information to generate tar-gets that should re-create the loudness rela-tionships between speech and warble tonesfound in individuals with normal hearing.As such, aided perceptions of warble toneand speech stimuli demonstrated by thehearing aid wearer should be similar tothose demonstrated by a group of individu-als with normal hearing.

Figure 1–7 provides an example of the tar-get data provided by VIOLA. Input/outputcurves are provided at frequencies desig-nated by the audiologist. A hearing aid pa-rameter table is provided to assist the audi-ologist in choosing appropriate hearing aids.

12843.C01.PGS 3/8/02 10:58 AM Page 14


The audiologist also may enter these para-meters from a given hearing aid, and the re-sulting output can be plotted on the input/output graphs to see how closely the targetis matched and where compromises may benecessary.

Comparison among Various NonlinearFitting Strategies

Fitting strategies employing an LE versus anLN approach would be expected to generatediffering prescriptions. Byrne et al (2001)

Figure 1–4. Example of prescriptive output provided by desired sensation levelinput/output (DSL [i/o]).

12843.C01.PGS 3/8/02 10:58 AM Page 15


found that the degree of difference amongseveral fitting strategies (DSL [i/o], FIG6,VIOLA, and NAL-NL1) varied with differ-ent degrees and configurations of hearingloss. For example, one would expect to findrelatively greater amounts of low-frequencyamplification using the LN rationale. Figure1–8 demonstrates this finding when com-

paring the REIG targets generated by FIG6and NAL-NL1 for a flat, 50-dB sensorineuralhearing loss. The NAL-NL1 prescriptiondemonstrates relatively less low-frequencygain and greater amounts of amplification inthe midfrequencies.

What is somewhat surprising is the degreeof difference found among fitting prescrip-

Figure 1–5. An example sound pressure level SPL-O-Gram generated using DSL [i/o]software.

12843.C01.PGS 3/8/02 10:58 AM Page 16


tions that incorporate a similar underlyingrationale. In these instances, the differencesare likely related to how a given rationale isimplemented. For example, NAL-NL1 andDSL [i/o] can prescribe a dramatically dif-ferent degree of high-frequency amplifica-tion even when DSL [i/o] is configured to

generate targets meant to result in loudnessequalization. Figure 1–9, for example, pro-vides REAG targets generated by DSL [i/o]and NAL-NL1 for a steeply sloping, high-frequency sensorineural hearing loss. DSL[i/o] prescribes a much greater amount ofhigh-frequency gain for soft and average

Figure 1–6. Example of prescriptive output provided by FIG6.

12843.C01.PGS 3/8/02 10:58 AM Page 17


input levels. Differences in how amplifica-tion is prescribed as a function of hearingloss slope and how severe degrees of hear-ing loss are treated help account for thesefindings.

Strategies that incorporate an LN rationalealso can differ substantially. Figure 1–10 pro-

vides coupler targets generated by VIOLAand FIG6 for a mild-to-moderate, gently slop-ing sensorineural hearing loss. Ricketts (1996)found substantial differences between pre-scriptions generated by various LN strategies.These differences were attributed to differingalgorithms, frequency response rules, and as-

Figure 1–7. Example of prescriptive output provided by the visual input-output locatoralgorithm (VIOLA).

12843.C01.PGS 3/8/02 10:58 AM Page 18


Figure 1–8. Insertion gain targets prescribed by FIG6 and NAL-NL1 for flat, 50-dB senso-rineural hearing loss (HL).

Figure 1–9. Real-ear-aided gain targets provided by DSL [i/o] and NAL-NL1 for a mild-to-severe, steeply sloping, high-frequency sensorineural hearing loss.

Figure 1–10. Coupler gain targets from VIOLA and FIG6 for a gently sloping, mild-to-moderate sensorineural hearing loss.

12843.C01.PGS 3/8/02 10:58 AM Page 19


sumptions regarding the signal processingemployed by the hearing aid (i.e., number ofindependent compression channels).

Which Method Is Best?

Considering the varying underlying theoriesand implementations employed with the fit-ting strategies outlined above, one might askwhether use of any particular strategy willyield more successful hearing aid fittings. In-tuitively, one might assume that restorationof normal loudness perception through useof patient-specific measures of dynamicrange has the greatest face validity and rep-resents a good starting point (Valente andVan Vliet, 1997). Unfortunately, there is scantempirical research that has addressed the rel-ative real-world performance provided byfitting strategies incorporating an LN ratio-nale versus strategies incorporating a differ-ent rationale (e.g., LE).

Several researchers have suggested thatthe loudness relations among various fre-quency regions are not important so long asspeech sounds are made audible, but not un-comfortably loud (Humes, 1996). This sug-gestion is based on the results from researchconducted with linear fitting strategies thathave shown that a fairly wide range of fre-quency responses can lead to similar speechrecognition ability and acceptable soundquality when audibility of the signal hasbeen controlled (Sullivan et al, 1988). Thegoal is to fit as much of the speech signal aspossible within the patient’s dynamic range.Other researchers have suggested that theloudness relations among various frequencyregions are important when prescribing fit-ting targets, and that restoring normal loud-ness relations may not be an appropriategoal (Byrne, 1996).

With an LN response, the lower frequen-cies are likely determining the overall usegain, and it is possible that the higher fre-quencies may be presented at a less thanideal level, especially if significant back-ground noise is present and/or soft speech isencountered. Although this is normal (i.e., anindividual with normal hearing would per-

ceive the higher frequency speech sounds ata low sensation level), such a response maynot be ideal for the patient with hearing losswho has a compromised auditory system. Inthis case, an LE approach may be better as itensures that the critical high-frequency re-gions contribute more to the overall loudnessof the signal (Byrne, 1996).

There also is empirical support for the as-sumption that restoration of normal loud-ness perception may not lead to the highestspeech recognition ability when predictedspeech recognition ability is compared viacalculations of the Articulation Index (AI) orthe SII (Rankovic, 1995; Ricketts, 1996; Stel-machowicz et al, 1998). The higher AI resultsfrom a greater amount of gain, most notablyin the higher frequencies, prescribed bystrategies that attempt to maximize intelligi-bility using threshold data.

Whether these same findings would re-main if actual speech recognition ability wasobtained is unknown. In addition, compari-son in this manner does not address differ-ences in perceived loudness and sound qual-ity that must be considered when comparingprescriptions (Studebaker, 1992). In the Rick-etts (1996) study, for example, differenceswere minimized once loudness was equal-ized. The question of whether the majorityof patients with hearing loss would tolerateor eventually adapt to the greater amount ofhigh-frequency amplification prescribed bysome strategies has not been answered(Lindley et al, 2001).

Summary

Prescriptive fitting strategies calculatethe desired electroacoustic characteristicsof a hearing aid based on the results of vari-ous psychoacoustic data such as pure-tonethresholds, MCL, LDL, and dynamic range.Currently, the audiologist has access to a va-riety of independent prescriptive formulasthat achieve selective amplification for agiven patient. Several of the prescriptive for-mulas now are flexible enough to providethe audiologist with the ability to use asmuch or as little psychoacoustic data as de-

12843.C01.PGS 3/8/02 10:58 AM Page 20


sired (e.g., NAL-NL1, DSL [i/o]) and allowthe audiologist to deal with both linear andnonlinear technology within the same pre-scriptive formula (e.g., DSL [i/o], VIOLA).The audiologist using advanced technologyis faced with a variety of proprietary fittingformulas that have not been well defined inthe clinical or research literature. A historicalperspective on prescriptive formulas and anunderstanding of the current independentfitting strategies should assist the audiolo-gist in evaluating proprietary fitting algo-rithms in the future.

References

American National Standards Institute. Methods forCalculation of the Speech Intelligibility Index (ANSIS2.5–1997). Accredited Standards Committee S3,Bioacoustics. Washington, DC: ANSI, 1997.

Berger K, Hagberg E, Rane R. Prescription of HearingAids: Rationale, Procedures and Results, 5th ed. Kent,OH: Herald, 1989.

Berger K, Soltisz L. Variability of thresholds and MCLswith speech babble. Aust J Audiol 1981;3:1–3.

Bragg V. Toward a more objective hearing aid fittingprocedure. Hear Instrum 1977;28(9):6–9.

Byrne D. Effects of frequency response characteristicson speech discrimination and perceived intelligibil-ity and pleasantness of speech for hearing-impairedlisteners. J Acoust Soc Am 1986;80:494–504.

Byrne D. Key issues in hearing aid selection and evalua-tion. J Am Acad Audiol 1992;3:67–80.

Byrne D. Hearing aid selection for the 1990s: Where to?J Am Acad Audiol 1996;7:377–395.

Byrne D, Cotton S. Preferred listening levels for sen-sorineurally hearing-impaired listeners. Ear Hear1987;9:7–14.

Byrne D, Dillon H. The National Acoustics Laborato-ries’ (NAL) new procedure for selecting the gain andfrequency response of a hearing aid. Ear Hear1986;7:257–265.

Byrne D, Dillon H, Ching T, Katsch R, Keidser G. NAL-NL1 procedure for fitting nonlinear hearing aids:characteristics and comparisons with other proce-dures. J Am Acad Audiol 2001;12:37–51.

Byrne D, Parkinson A, Newall P. Hearing aid gain andfrequency response requirements for the severely/profoundly hearing impaired. Ear Hear 1990;11:40–49.

Carhart R. Tests for the selection of hearing aids. Laryn-goscope 1946;56:780–794.

Ching T, Dillon H, Byrne D. Speech recognition of hear-ing-impaired listeners: predictions from audibilityand the limited role of high-frequency amplification.J Acoust Soc Am 1998;103:1128–1140.

Cornelisse L, Seewald R, Jamieson D. The input/outputformula: a theoretical approach to the fitting of per-sonal amplification devices. J Acoust Soc Am 1995;97(3):1854–1864.

Cox R. The MSUv3 hearing instrument prescriptionprocedure. Hear Instrum 1988;39(1):6–10.

Cox R, Alexander G, Taylor I, Gray G. The Contour Testof Loudness Perception. Ear Hear 1997;18;388–400.

Cox R, Bisset J. Prediction of aided preferred listeninglevels for hearing aid gain prescription. Ear Hear1982;3:66–71.

Cox R, Flamme G. Accuracy of predicted ear canalspeech levels using the VIOLA input/output-basedfitting strategy. Ear Hear 1998;19:139–148.

Davis H, Hudgins C, Marquis R, et al. The selection ofhearing aids. Layrngoscope 1946;56:85–115,135–163.

Gitles T, Niquette P. FIG6 in ten. Hear Rev 1995;2:28–30.Hogan C, Turner C. High frequency audibility: Benefits

for hearing-impaired listeners. J Acoust Soc Am 1998;104:432–441.

Humes L. An evaluation of several rationales for select-ing hearing aid gain. J Speech Hear Dis 1986;51:272–281.

Humes L. And the winner is . . . Hear Instrum 1988;39:24–26.

Humes L. Evolution of prescriptive fitting approaches.Am J Audiol 1996;5:19–23.

Kamm C, Dirks D, Mickey R. Effect of sensorineuralhearing loss on loudness discomfort level. J SpeechHear Res 1978;21:668–681.

Killion M. Recent earmolds for wideband OTE and ITEhearing aids. Hear J 1981;37(8):15–22.

Levitt H, Sullivan J, Neuman A, Rubin-Spitz J. Experi-ments with a programmable master hearing aid.J Rehab Res Dev 1987;25:29–54.

Libby E. The 1/3–2/3 insertion gain hearing aid selec-tion guide. Hear Instrum 1986;37:27–28.

Lindley G, Palmer C, Durrant J, Pratt S. Audiologistversus patient driven hearing aid fitting protocols.Semin Hear 2001;22:139–160.

Lybarger S. U.S. Patent Application SN 532, 278, 1944. McCandless G, Lyregaard P. Prescription of gain and

output (POGO) for hearing aids. Hear Instrum 1983;3:16–21.

Palmer C, Lindley G. Reliability of the contour test in apopulation of adults with hearing loss. J Am AcadAudiol 1998;9:289–215.

Pascoe D. Frequency responses of hearing aids andtheir effects on the speech perception of hearing-impaired subjects. Ann Otol Rhinol Laryngol 1975;84(suppl 23):1–40.

Pascoe D. Clinical measurements of the auditory dy-namic range and their relation to formulas for hear-ing aid gain, in presbyacusis and other age relatedaspects. 14th Danavox Symposium 1990;129–147.

Rankovic CM. Derivation of frequency-gain character-istics for maximizing speech reception in noise. JSpeech Hear Res 1995;38:913–929.

Ricketts T. Fitting hearing aids to individual loudness-perception measures. Ear Hear 1996:17;124–132.

Schwartz D, Lyregaard P, Lundh P. Hearing aid selec-tion for severe to profound hearing loss. Hear J1988;41(2):13–17.

Schwartz D, Walden B. Current status of the clinicalhearing aid evaluation. In: Studies in the Use of Am-plification for the Hearing Impaired: Proceedings of aSymposium. Copenhagen: Excerpta Medica, 1980,pp. 15–28.

Seewald R. The desired sensation level method for fit-ting children: version 3.0. Hear J 1992;45(4):36–41.

12843.C01.PGS 3/8/02 10:58 AM Page 21


Seewald R. Fitting children with the DSL method. Hear J1994;47(9):10,48–51.

Seewald R, Cornelisse L, Ramji K, Sinclair S, Mookie K,Jamieson D. DSL v4.1 for Windows: A Software Imple-mentation of the Desired Sensation Level (DSL [i/o])Method for Fitting Linear Gain and Wide-Dynamic-Range Compression Hearing Instruments. London, On-tario: University of Western Ontario, 1997.

Seewald R, Ross M, Spiro M. Selecting amplificationcharacteristics for young hearing-impaired children.Ear Hear 1995;6:48–53.

Shapiro I. Hearing aid fitting by prescription. Audiology1976;15:163–173.

Shapiro I. Comparison of three hearing aid prescriptionprocedures. Ear Hear 1980;1:211–214.

Skinner M. Speech intelligibility in noise-induced hear-ing loss: effects of high frequency compensation. JAcoust Soc Am 1980;67:306–317.

Skinner M. Hearing Aid Evaluation. Englewood Cliffs,NJ: Prentice-Hall, 1988.

Skinner M, Karstaedt M, Miller J. Amplification band-width and speech intelligibility for two listenerswith sensorineural hearing loss. Audiology 1982a;21:251–268.

Skinner M, Miller J. Amplification bandwidth and intel-ligibility of speech in quiet and noise for listenerswith sensorineural hearing loss. Audiology 1983;22:253–279.

Skinner M, Pascoe D, Miller J, Popelka G. Measurementsto determine the optimal placement of speech energy

within the listener’s auditory area: a basis for select-ing amplification characteristics. In: Studebaker G,Bess F, eds. The Vanderbilt Hearing-Aid Report. UpperDarby, PA: Monographs in Contemporary Audiol-ogy 1982b, pp. 161–169.

Stelmachowicz GP, Dalzell S, Peterson D, Kopun J,Lewis DL, Hoover BE. Comparison of threshold-based fitting strategies for nonlinear hearing aids.Ear Hear 1998;19(2):131–138.

Stephens S, Blegvad B, Krogh H. The value of somesuprathreshold auditory measures. Scand Audiol1977;6:213–221.

Studebaker G. The effect of equating loudness on audi-bility-based hearing aid selection procedures. J AmAcad Audiol 1992;3:113–118.

Sullivan J, Levitt H, Hwang J, Hennessey A. An experi-mental comparison of four hearing aid prescriptionmethods. Ear Hear 1988;9:22–32.

Thorton A, Raffin M. Speech-discrimination scoresmodeled as a binomial variable. J Speech Hear Dis1978;21:507–518.

Valente M, Van Vliet D. The independent hearing aidfitting forum (IHAFF) protocol. Trends Ampl 1997;2:6–35.

Walden B, Schwartz D, Williams D, Holum-HardegenL, Crowley J. Test of the assumptions underlyingcomparative hearing aid evaluations. J Speech HearDis 1983;48:264–273.

Watson N, Knudsen V. Selective amplification in hear-ing aids. J Acoust Soc Am 1940;11:406–419.

12843.C01.PGS 3/8/02 10:58 AM Page 22

Documents

Overview and Rationale for Prescriptive Formulas …1 Overview and Rationale for Prescriptive Formulas for Linear and Nonlinear Hearing Aids CATHERINE V. PALMER, GEORGE A. LINDLEY