German Oldenburg Sentence Test for Children: A

Original Paper Folia Phoniatr Logop 2012;64:227–233 DOI: 10.1159/000342414

Published online: September 28, 2012

German Oldenburg Sentence Test for Children: A Useful Speech Audiometry Tool for Hearing-Impaired Children at Kindergarten and School Age Tobias Weißgerber a, b Uwe Baumann b Thomas Brand c Katrin Neumann a, d a

Department of Phoniatrics and Pediatric Audiology and b Audiological Acoustics, ENT Clinic, Goethe University, Frankfurt am Main, c Medical Physics Section, Carl von Ossietzky University, Oldenburg, and d Phoniatrics and Pediatric Audiology, St. Josef and St. Elisabeth Hospital, Ruhr-Universität Bochum, Germany

Key Words Speech audiometry ⴢ Children ⴢ Oldenburg Sentence Test for Children ⴢ Discrimination function

Abstract Background/Aims: In speech audiometry, sentence tests have the advantage of assessing more words within a given period of time than single-word tests do. Consequently, greater accuracy (steeper discrimination function) is achieved. The recently developed German Oldenburg Sentence Test for Children (OlKiSa) has been evaluated thus far for school-aged children in noise, and normative data for younger children in a quiet environment have been established. In this study, the focus is on its applicability in hearing-impaired children fitted with hearing aids or cochlear implants. Methods: The use of the OlKiSa in a quiet environment in hearing-impaired children aged 4 years or older was evaluated. One hundred and nineteen hearing-impaired children aged between 4 and 10 years performed the OlKiSa in a quiet environment. Individual speech reception thresholds (SRT) were measured aided and unaided and the slopes of the discrimination functions were calculated. Results: Independent of age, the mean slope of the discrimination function for SRT was about 7%/dB in both conditions, which

© 2012 S. Karger AG, Basel 1021–7762/12/0645–0227$38.00/0 Fax +41 61 306 12 34 E-Mail [email protected] www.karger.com

Accessible online at: www.karger.com/fpl

is a high value of steepness for a speech audiometric test in a quiet environment. Conclusion: The OlKiSa in quiet is a reliable test procedure for hearing-impaired children aged 4 years and older. Copyright © 2012 S. Karger AG, Basel

Introduction

Reliable audiological test procedures are mandatory in order to detect and treat hearing impairments in children. The main goal of therapy for hearing-impaired children is to facilitate age-appropriate speech and language development. Common subjective pediatric audiometric methods such as aided free field audiometry only estimate the perception threshold of non-speech stimuli. Therefore, speech perception tests for children are of special interest. Speech audiometry in children should rather be measured at speech levels high above the individual threshold [1] so that essential information required for the proper fitting of hearing aids or cochlear implants (CIs) can be obtained. Speech in everyday life situations is often masked by background noise. Speech audiometry in noise, in turn, is the best approach to replicate this hearing situation. Hence, it is the most widely used test Tobias Weißgerber Audiological Acoustics, ENT Department, Goethe University Theodor-Stern-Kai 7, House 8 DE–60590 Frankfurt am Main (Germany) Tel. +49 69 6301 5898, E-Mail tobias.weissgerber @ kgu.de

procedure in the rehabilitation of children with hearing disorders. However, for children whose linguistic abilities in the lingua franca are limited due to hearing loss, migration background, or a developmental language disorder, an assessment of speech perception is sometimes only feasible in a quiet environment. Therefore, reliable speech audiometric tests for testing children in quiet are also needed. This study focuses on speech tests in quiet for hearing-impaired children. Speech audiometric tests for children have to be ageappropriate with respect to cognition, attention, and motor development [2]. The task has to be interesting, motivating, short, and time-efficient. The test material should correspond to the speech perception abilities of a child, use age-appropriate vocabulary, take phonetic/phonological errors into consideration, be phonetically balanced, and be presented on standardized media. Additionally, the test has to fulfil the main quality criteria of psychometric tests and to consider learning effects. Finally, test lists have to be homogeneous. Discrimination Function of a Speech Test The listener’s speech intelligibility/discrimination as a function of speech level is described by the discrimination function. Speech intelligibility is defined by word scoring, which means the mean probability p that the words of a sentence are repeated correctly by the listener. Alternatively p is defined as the probability of repeating whole sentences correctly (sentence scoring). For speech tests performed in quiet, speech level L is the sound pressure level of the speech signal. In noise, L is the signal-tonoise ratio. A logistic function represents the discrimination function (equation 1). L50 denotes the speech reception threshold (SRT), which refers to a 50% probability of correct responses. The parameter s denotes the slope of the discrimination function at L50 [3]. p (L, SRT , s) =

1

1 + exp(4¸ s¸(L50 - L))

(1)

The measuring accuracy of a speech audiometric test is defined by the slope of the discrimination function at the SRT. The slope should be high; its doubling elicits a doubling of precision. Hence, several words per test should be presented, because an increase in the word number by factor n decreases the standard deviation by factor 冪n [3]. Commonly used German speech audiometric tests for children, for example, the Mainz Speech Test for Children (Mainzer Kindersprachtest) [4], the Goettingen 228

Folia Phoniatr Logop 2012;64:227–233

Child Test for Speech Perception (Göttinger Kindersprachverständnistest) [5, 6] and the Oldenburg Children’s Rhyme Test (Oldenburger Kinderreimtest, OlKi) [7, 8], use single words. Consequently, in order to obtain sufficient reliability, a long test phase is necessary. However, children’s auditory span is short, maturing only with increasing age; it encompasses less than five words for children at the age of 6 years [9]. There are, additionally, other pitfalls with the abovenamed audiometric tests. The linguistic material of the Mainz Speech Test for Children is not phonemically balanced. The Goettingen Child Test for Speech Perception tests only monosyllables and has a relatively low discrimination function slope. The same holds for the OlKi. Furthermore, none of the aforementioned tests is suitable for determining the SRT in noise because of insufficient steepness of their discrimination functions. Given that the duration of speech audiometric tests is limited by the early weariness of young children, procedures which present more words per time unit than a single-word test (a so-called ‘sentence test’) are expected to increase the reliability of test results. Nevertheless, there are many additional points to consider. The intelligibility of speech is related to utterance length and is nearly independent (at least for normal-hearing subjects) of the average rate of speaking [10]. However, the speech intelligibility of CI recipients decreases often with increasing rate of speaking. The slope of the discrimination function of the sentence test is determined by the slope of the single words, the predictability j of the speech material [11], the number of trials and the type of adaptive test algorithm. Adaptive testing procedures have become popular in psychophysical experiments over the past 20 years due to their efficiency and speed [12]. The level of a stimulus on each experimental trial is adaptively determined by performance on previous trials. These methods converge rapidly on a given level of performance and concentrate experimental trials in the range of the measurement of interest. Oldenburg Sentence Test for Children The adaptive German Oldenburg Sentence Test (Oldenburger Satztest, OlSa) [13–15], for schoolchildren and adults alike and consisting of five-word sentences, was derived from the Swedish sentence test by Hagerman [16]. It has been shown to be a reliable and valid test procedure for adults with only small training effects, a high slope of the discrimination function and very low predictability [15]. However, test results obtained with elementary school children were less consistent than those for adults [17]. One reason for this might be that the auditory span Weißgerber /Baumann /Brand /Neumann

of children in the first grade is often still too restricted for application of the OlSa. Consequently, the Oldenburg Sentence Test for Children (Oldenburger Kindersatztest, OlKiSa) [17] was derived from the material of the OlSa and designed to be applicable for children of kindergarten age and upwards. The OlKiSa presents pseudo-sentences consisting of three words with a numeral, an adjective, and an object (e.g. ‘four red flowers’). The OlKiSa has the same qualities as the OlSa with respect to low predictability and phonetically balanced speech material. The OlKiSa was first evaluated solely in noise with normal-hearing children and showed higher homogeneity than the OlSa. Preliminary results of the clinical application of this test in quiet and in noise showed that the procedure seemed to be applicable for children as young as 4 years of age [18]. Neumann et al. [19] derived age-dependent reference functions for the SRT of normal-hearing children aged from 4 to 9 years in quiet and showed that the slope of the discrimination functions of the OlKiSa was higher when compared to all other common pediatric speech audiometric tests in Germany. Validity, reliability, and objectivity values of the test were also excellent. Training effects were overcome with two training lists before the main testing. Until recently, the OlKiSa was only evaluated in a large group of normal-hearing children [19]. These findings are not necessarily valid for children with hearing loss. It is known that people with hearing loss possess a shallower discrimination function [3]. Therefore, additional evaluation with hearing-impaired children is necessary to verify whether the OlKiSa is also a precise and reliable test for these patients. Ebner et al. [20] conducted tests in a small group of deaf children in the aided CI condition with very limited generalizability of these findings. The aim of this work is (1) to investigate the applicability of the OlKiSa in quiet for diagnostic and therapy control purposes in hearing-impaired children with various types and degrees of hearing loss; (2) to assess the slopes of the discrimination functions, and (3) to investigate the relationship of the SRT with pure-tone audiometry.

only a mild hearing loss in the better ear) or bilateral hearing aid or CI or were bimodal (hearing aid on one side and CI on the other). Medical conditions other than hearing loss, in particular intellectual and multiple disabilities or syndromes, were ruled out by a trained ear, nose and throat physician/phoniatrician/pediatric audiologist. The sample included 84 monolingual German children and 35 children with non-German native language. The children of the latter group were pretested by the examiner by presenting pseudo-sentences of the OlKiSa speech material at about 65 dB SPL. If they perceived the sentences correctly while using their hearing instruments, further measurements were conducted. Initially, binaural tympanometry was performed. Furthermore, unaided pure-tone audiometry (500, 1,000, 2,000, 4,000 Hz) with headphones as well as aided free field audiometry were conducted to calculate the pure-tone average (PTA) for both aided and unaided conditions. Three children with hearing impairment were not fitted with hearing aids at the date of measurement and were therefore only tested unaided. The children with CI were only measured aided. Twenty-three of the 119 children were excluded from the study due to insufficient speech perception or production skills for developmental or compliance reasons. Most of the excluded participants were children (of all age groups) at schools for the hard of hearing with congenital deafness and cochlear implantation after the age of 2 years. The measurements of the remaining 96 children were included in the data analysis. The PTA values ranged from 14 to 88 dB HL (unaided) and from 22 to 40 dB HL (aided) across age classes. The aided and unaided hearing thresholds of the age groups from 6 to 9 years were statistically comparable. The age groups 4, 5 and 10 years were not included in the statistical comparison between age groups due to insufficient sample size. The children’s age, age-related sample sizes, gender, PTA, and status of hearing device fitting are documented in table 1. The children were tested by the first author in one of the following locations: (1) the clinical audiological setting of the author’s department, (2) three regional schools for the hard of hearing, and (3) a hearing aid retailer. The study was conducted according to the principles of the Helsinki Accords and all parents provided written consent.

Participants One hundred and nineteen children (median age: 7 years, range: 4–10 years), 64 male, 55 female, with monaural or binaural conductive or sensorineural hearing loss participated in this study. The children were fitted with either a unilateral (in case of

Study Design The OlKiSa is a commercially available tool implemented as part of the software Oldenburg Measurement Applications© by HörTech and can be applied on any standard personal computer. The test includes 20 test lists with 7 items each and 10 double lists with 14 items each. To convert the sound signals into the analog domain the DA converter ‘Multiface II’ by RME was used. The playback level via headphones was calibrated with a coupling device (Brüel & Kjaer, type 4153), playback via loudspeakers in free field with a sound pressure level meter. All measurements were taken in a sound booth with the exception of the measurements at one regional school for the hard of hearing, where the tests were performed in a room with a short reverberation time and low background noise. The A-weighted equivalent sound pressure level with time constant ‘fast’ in this testing room was measured with a sound pressure level meter and did not exceed 35 dB SPL. The calibration of the sound pressure level in the free field was checked and readjusted at all different locations of measurement. In order to ensure high objectivity, all children obtained the same test instructions.

Oldenburg Sentence Test for Children


Methods

229

Table 1. Participants’ age, sample size n, gender, aided and unaided PTA and status of fitting with hearing devices (hearing aid, CI or no hearing device)

Age, years

n Gender, m/f PTAunaided, dB HL PTAaided, dB HL Hearing aid/CI/–

4

5

6

7

8

9

10

5 3/2 51837 3585 4/0/1

6 5/1 50816 2682 6/0/0

18 8/10 41821 3085 14/4/0

23 12/11 44821 2885 20/2/1

19 11/8 51822 2886 14/4/1

21 13/8 35818 2885 16/5/0

4 2/2 43811 2883 3/1/0

Procedure Training occurred prior to testing by presenting the test items via loudspeaker to the children while wearing their hearing devices. To become familiar with the procedure, participants were initially presented with a test list consisting of 7 items at a fixed sound pressure level that was 30 dB higher than the subject’s individual PTA. A second list with another 7 items was subsequently presented and the SRT was measured adaptively. No further training effects were expected to occur after training [19]. The measurements were taken in aided (all hearing devices, signal presentation via loudspeaker) and unaided (monaural signal presentation via closed circumaural headphones, Sennheiser HDA200) conditions. In the latter situation, only the ear with the better PTA was tested to avoid weariness. For each condition, two test lists of 14 items each were conducted and the mean SRT was calculated from the adaptively estimated single SRTs. Two other test lists were used in addition to determine the percentage of perceived words 2 dB above and 2 dB below the previously measured SRT. Adaptive Measurement of the SRT The adaptive procedure to estimate the SRT was based on the method developed by Hagerman and Kinnefors [21, 22] and modified by Brand [23]. The discrimination value of the previously presented sentence is used as input to the adaptive level ⌬L setting according to the equation

⌬L = -

f (i )¸( prev - tar )

, s in which the parameter tar denotes the target discrimination value at which the procedure should converge (50%), prev denotes the discrimination value obtained in the previous sentence, s denotes the slope. The parameter f(i) controls the rate of convergence [3]. f(i) equals 2 at the start of the measurement, 1 after the first reversal and 0.5 after the second reversal. The starting level was 30 dB above the particular pure-tone threshold.

Estimation of the Discrimination Function With the three values SRT and speech perception at sound pressure levels of SRT + 2 dB and SRT – 2 dB, respectively, the slope of the discrimination function in the SRT data point was estimated by logistic regression. In order to calculate the normative values for the normal-hearing children, age-related discrimination functions

230


were fitted into the averaged individual SRT values and the percentages of perceived words at SRT + 2 dB and SRT – 2 dB [19]. In contrast, normative data for the hearing-impaired children in this work were calculated separately for each subject using the slopes of the discrimination functions because of the large interindividual variance of the SRTs due to the different degrees of hearing loss.

Results

SRT and Pure-Tone Audiometry The results of the SRT measurements for all age groups are given in table 2. Due to the different degrees of hearing loss across participants it was not possible to compare the values of the SRTunaided between the different age groups. In contrast, given that there is no significant difference in aided free field audiometry, statistical analysis of the SRTaided was feasible. As shown by Neumann et al. [19], no further learning effect could be found after conducting one training list. The SRTaided marginally but significantly correlated with aided free field audiometry (r = 0.35, p ! 0.005) and SRTunaided strongly correlated with the PTA (r = 0.87, p ! 0.005). Slope of the Discrimination Function The slopes of the discrimination functions s were calculated separately for every participant and averaged for each age group. The mean values and standard deviations of the age-dependent slopes in unaided and aided conditions are shown in table 2. Without hearing devices, the slopes of the discrimination functions in the SRT ranged between approximately 4.7 and 8.8%/dB, and in aided conditions between 6.1 and 7.8%/dB. The slopes of the unaided and aided conditions are statistically comparable. The same holds for the mean slopes of CI (5.13 8 2.75%/dB) and hearing aid recipients (6.27 8 3.84%/dB). Weißgerber /Baumann /Brand /Neumann

100 90 80

SRT (dB SPL)

70 60 50 40 30 20 10 0 0

Fig. 1. Correlation of unaided SRT and un-

10

20

30

40

50

60

70

80

90

100

PTA (dB HL)

aided PTA (0.5, 1, 2, 4 kHz), averaged over all age groups.

Table 2. Mean SRTs and mean slopes of the discrimination functions (81 SD) for all age groups in unaided and aided conditions

Age, years 4 SRTunaided, dB SPL SRTaided, dB SPL sunaided, %/dB saided, %/dB

5

6

7

8

9

10

mean

57.51827.7 53.4823.1 47.94817.9 55.54821.0 57.57820.8 45.24818.2 56.6587.1 53.4819.6 41.63814.5 31.3886.1 38.85810.2 38.3187.1 37.21810.1 35.1687.4 37.7588.7 37.288.7 4.7781.2 5.7382.3 6.1284.3 6.3282.7 8.3382.7 7.5684.2 8.7883.1 6.883.4 6.9586.3 6.9484.4 6.0983.1 6.7583.4 7.4982.7 6.6982.9 7.7581.7 783.2

Discussion

Estimation of SRT and Steepness s In this study, only the SRT was measured adaptively, whereas the steepness s was manually calculated from the SRT and speech perception 2 dB above and below the SRT. Another possibility would have been to estimate the slope of the SRT concurrently with the SRT in a single measurement. This algorithm is implemented in the Oldenburg Measurement Applications©. However, in this approach at least 30 sentences need to be tested in order to obtain reliable results for the slope estimation and the SRT measurement [3]. Furthermore, the adaptive estimation of both SRT and s tends to systematically overestimate s and could lead to incorrect s values [24].

sa for both aided and unaided audiometry for diagnostic reasons. Furthermore, the amplification of the hearing devices is not linear and acts differently at hearing threshold than at the level of the SRT. However, the unaided SRT could give at least a rough estimate of the PTA and vice versa, at least for higher degrees of hearing loss (fig. 1), as the difference between PTA and SRT decreases with ascending PTA. For the fitting of CIs or hearing aids and to measure the benefit of different hearing devices, the measurement of both PTA and SRT is mandatory.

Relationship of PTA and Speech Test The PTA of hearing-impaired children is not sufficient to estimate speech perception in quiet and vice ver-

Hearing-Impaired and Normal-Hearing Children Neumann et al. [19] measured the SRTs and estimated s for normal-hearing children in the same age groups reported in the current study, and provided age-dependent normal ranges for the SRT in quiet. The mean slope of all age groups was found to be approximately 10%/dB. For hearing-impaired children, not only were the SRTs higher than those in normal-hearing subjects, but also the



231

slopes were about 43% lower in both unaided with headphones and aided in free field conditions. Ebner et al. [20] reported a similar result of about 50% decrement in the slopes of CI patients compared to normal-hearing subjects. Typical values for the slope of the discrimination function of well-designed sentence tests in noise of adults are between 20%/dB for normal-hearing listeners and 5%/dB for severely hearing-impaired listeners [3]. Hence, the mean slope of about 7%/dB of the OlKiSa in quiet is comparatively high for hearing-impaired children and qualifies this speech test for measurement in clinical routine. The OlKiSa in quiet can be used with hearing aid and CI recipients with the same reliability. Comparison of the OlKiSa with Other German Speech Tests The mean steepness of the OlKiSa in quiet with hearing-impaired children exceeds even the steepness of many single-word tests for normal-hearing children. Thus, this sentence test is better suited for clinical practice than a single-word test [17]. The single-word test OlKi (performed on normal-hearing schoolchildren in noise), for example, has a reported steepness of only 5%/dB [17]. Sentence tests generally have shown a higher steepness in noise than in quiet. In order to assess the benefit of hearing devices, SRT measurements in noise should be preferred over measurements in quiet, especially given that children are able to comprehend the task of testing speech perception in noise. If only a test in quiet is applicable, the OlKiSa is the most precise German speech test. For spe-

cial diagnostic purposes, as in discrimination of minimal phonological differences or central auditory processing disorders, a single-word test may be preferable.

Conclusions

The OlKiSa is a reliable speech test in noise (previous studies) as well as in quiet (present study). Speech perception can be assessed in younger children with comparable precision to adults and with the advantage of a higher sensitivity compared to single-word tests using the OlKiSa. The OlKiSa is a reliable tool not only in the diagnosis of hearing impairment but also for measuring the benefit of fitting a hearing aid or CI. We, therefore, recommend that this test be used in daily clinical practice instead of the still commonly used German single-word tests.

Acknowledgements The authors would like to thank Cochlear Deutschland GmbH and Phonak AG for their financial support. Parts of this work were presented at the 13th Annual Congress of the German Audiology Association (Deutsche Gesellschaft für Audiologie) in Frankfurt am Main, Germany, March 17–20, 2010.

Disclosure Statement There are no conflicts of interest.

References 1 Mendel LL: Current considerations in pediatric speech audiometry. Int J Audiol 2008; 47:546–553. 2 Kosky C, Boothroyd A: Validation of an online implementation of the imitative test of speech pattern contrast perception (IMSPAC). J Am Acad Audiol 2003;14:72–83. 3 Brand T, Kollmeier B: Efficient adaptive procedures for threshold and concurrent slope estimates for psychophysics and speech intelligibility tests. J Acoust Soc Am 2002;111: 2801–2810. 4 Biesalski P, Leitner H, Leitner E, Gangel D: Der Mainzer Kindersprachtest, Sprachaudiometrie im Vorschulalter. HNO 1974; 22:160–161.

232

5 Chilla R, Gabriel P, Kozielski P, Bänsch D, Kabas M: Der Göttinger Kindersprachverständnistest. I. Sprachaudiometrie des Kindergarten- und retardierten Kindes mit einem Einsilber-Bildtest. HNO 1976; 24: 342– 346. 6 Gabriel P, Chilla R, Kiese C, Kabas M, Bänsch D: Der Göttinger Kindersprachverständnistest II. HNO 1976;24:399–402. 7 Kliem K, Kollmeier B: Überlegungen zur Entwicklung eines Zweisilber-Kinder-Reimtests für die klinische Audiologie. Audiol Akustik 1995;34:6–10. 8 Brand T, Achtzehn J, Kollmeier B: Erstellung von Testlisten für den Oldenburger KinderReimtest. Z Audiol 1999;(suppl II):50–51. 9 Case R, Kurland DM, Goldberg J: Operational efficiency and the growth of short-term memory span. J Exp Child Psychol 1982; 33: 386–404.


10 Pollack I, Pickett JM: The Intelligibility of Excerpts from Conversation. Lang Speech 1963;6:165–171. 11 Boothroyd A, Nittrouer S: Mathematical treatment of context effects in phoneme and word recognition. J Acoust Soc Am 1988; 84: 101–114. 12 Leek MR: Adaptive procedures in psychophysical research. Percept Psychophys 2001; 63:1279–1292. 13 Wagener K, Kühnel V, Kollmeier B: Entwicklung und Evaluation eines Satztests für die deutsche Sprache I: Design des Oldenburger Satztests. Z Audiol 1999;38:4–15. 14 Wagener K, Brand T, Kollmeier B: Entwicklung und Evaluation eines Satztests für die deutsche Sprache II: Optimierung des Oldenburger Satztests. Z Audiol 1999; 38: 44– 56.

Weißgerber /Baumann /Brand /Neumann

15 Wagener K, Brand T, Kollmeier B: Entwicklung und Evaluation eines Satztests für die deutsche Sprache III: Evaluation des Oldenburger Satztests. Z Audiol 1999;38:4–15. 16 Hagerman B: Sentences for testing speech intelligibility in noise. Scand Audiol 1982; 11: 79–87. 17 Wagener K, Kollmeier B: Evaluation des Oldenburger Satztests mit Kindern und Oldenburger Kinder-Satztest. Z Audiol 2005; 44: 134–143. 18 Stephan K, Muigg F: Erfahrungen zur klinischen Anwendbarkeit des Oldenburger Kindersatztests (OLKISA). 12th Annu Conf Deutsche Ges für Audiol, Kiel, March 2008. Conference CD. ISBN 3–9809869–7–7.


19 Neumann K, Baumeister N, Baumann U, Sick U, Euler H, Weissgerber T: Speech audiometry in quiet with the Oldenburg Sentence Test for Children. Int J Audiol 2012; 51: 157– 163. 20 Ebner K, Steffens T, Hellbrück J: Sprachverstehen in Ruhe und im Störgeräusch und Lerneffekte bei normalhörenden und unilateral sowie sequentiell bilateral kochleaimplantierten Kindern. Z Audiol 2008; 47: 100–110.

21 Hagerman B, Kinnefors C: Efficient adaptive methods for measurements of speech reception thresholds in quiet and in noise. Karolinska Institutet, Teknisk Audiologi, Stockholm, 1993. 22 Hagerman B, Kinnefors C: Efficient adaptive methods for measurements of speech reception thresholds in quiet and in noise. Scand Audiol 1995;24:71–77. 23 Brand T: Effiziente Bestimmung psychometrischer Funktionen mit Sprachverständlichkeitstests; diploma thesis Georg August University Göttingen, 1994. 24 Leek MR, Hanna TE, Marshall L: Estimation of psychometric functions from adaptive tracking procedures. Percept Psychophys 1992;51:247–256.


233