Speech Audiometry Thresholds, Recognition Tests, MCL &UCL
Testing with Recorded Speech • 1000 Hz Calibration tone provided on recording • Balance calibration tone on VU meter at 0
Live Voice Testing: • Controlled Vocal Effort, • Adjust microphone sensitivity • to have the speech balanced at 0 dB on VU meter
Thresholds for Speech: • Speech Detection Threshold: lowest level at which the listener can tell that something is there (when the signal happens to be speech). • Speech Recognition Threshold: lowest level at which the listener can actual identify what the speech stimulus is.
SRT Stimuli: Spondees • 2-syllable words with equal stress • can be divided into two monosyllables • e.g., hotdog, baseball, whitewash, mousetrap, birthday, eardrum...
Why spondees? • because their intelligibility curves rise from near chance to 100% performance within a few decibels. • This provides a much more accurate threshold.
ASHA: Instruct Familiarize (16 words) Find Starting Level Begin Descending procedure Thresh = Start Level - # C + Corr. Factor Adaptive Instruct Familiarize Bracket (2-4 wds/level) Thresh = lowest level where you get at least 50% correctly repeated. SRT Methods
The SRT should agree with the PTA • (-8 to +6 dB of best threshold at these frequencies). • This way we have a double-check on results.
Exception for steeply-sloped high frequency losses. • SRT will be better than PTA • In this case use Fletcher average: 500 Hz + 1000 Hz / 2.
Masking for SRT's: • IF SRT - IA > best BC THRESH in NTE. • WHITE or PINK NOISE • STARTING LEVEL=SRTTE –35 +ABGNTE
Long Term Spectrum of Speech And the variance around it.
Bone Conduction SRT's: • useful in children • useful in malingerers.
Masking for SRT • If SRT - IA > best BC Thresh NTE • Put in at least: STARTING LEVEL=SRTTE –35 +ABGNTE But no more than: OVERMASK = EMNTE-IA> Best BC Thresh TE
Most Comfortable Loudness Level • Instructions important: you can strongly influence how a person responds. • "I am going to continue talking to you as I make my voice louder and softer. I will keep asking you to tell me whether my voice is too soft, too loud or comfortably loud." • Do a number of sweeps in level. • normally between 40 and 55 dB above SRT
Uncomfortable Loudness Level • Begin at MCL, raise level as you continue to talk. • "I am now going to ask you to tell me how my voice sounds to you as I make it louder. Please tell me if the level is comfortable, a little loud, or uncomfortably loud.” • Uncomfortable = loud enough so you would not want to listen to my voice for a long time.
Range of Comfortable Loudness • (Or the Dynamic Range for Speech) • = UCL – SRT • Normally 100 dB or greater • Unchanged in conductive losses • Can be much smaller in sensorineural hearing loss
Word Recognition Testing • Open set-client can respond with any word he/she can think of. • Closed set-response options are provided for the client (multiple choice test). • Free response-client is free to respond or not. • Forced Response-client must say something. • [Forced choice = closed set forced response.]
Phonetically Balanced Word Lists • selection of a group of words so that each phoneme appears with the same frequency it has in the normal lexicon. Based on Thorndike-Lorge lists of words and word frequencies. • So-called PB word lists-- CID W-22 Lists • Four lists of 50 words each.
CNC Word Lists • Consonant-Nucleus-Consonant words • Phonemically balanced • Four 50-item lists: the NU-6 Word Lists
Alternative Speech Choices • High Frequency Word Lists • Gardner’s Hi Frequency Word Lists • California Consonant Test • Nonsense Syllable Lists • The Nonsense Syllable Test (NST) • Sentence Tests • The Synthetic Sentence Identification (SSI) test • Speech Perception in Noise (SPIN) test • Connected Speech Test (CST) • The Hearing In Noise Test (HINT)
Children’s Tests • Word Intelligibility by Picture Identification (WIPI) test – six pictures to choose from. • Northwestern University Children’s Perception of Speech (NUCHIPS) test – four pictures to choose from • Monosyllable-Spondee-Trochee test – distinguishing word shapes, not identifying particular words; used in cochlear implant cases.
Administration: once you've chosen materials • transducer: AC, BC, or sound field. • level: at some SL (40)--may be specified by test. • response: repeat, write, or mark on list. • competing noise: what, how loud. • masking: calculate (Test Level - IA) - BC nte
Adaptive Testing • Finding a threshold, not a % score • Typically in presence of competing noise • Measure has low variance • Not yet common in clinical practice • (although we will talk about the HINT soon)
Performance-Intensity Functions • PI function: word recognition scores obtained at a range of stimulus levels. • Curve reaches a peak (Pbmax), and then • Either remains high (normal), or • Drops at higher levels (Rollover) • Rollover Index = (PBmax – Pbmin)/PBmax
Rollover Indices for the preceding examples • Normal: (100 - 100) / 100 = 0.0 • Rollover: (44 - 20) / 44 = 0.54 • Cochlear: (80 - 70)/80 = 0.125 • Rollover Indices of 0.45 or greater indicate a neural (VIIIth nerve) problem.
Cross Hearing & the Need to Mask • If Word level (HL)TE – IA > Best BCNTE • Use Pink (Speech) Noise, or white noise • Suggested Levels: EM = PBHL TE – IA + ABGNTE + 20 dB
Reliability of Word Recognition Scores • Range of possible scores upon retest Thornton & Raffin (1978)
NU-6 Words Ordered by Difficulty • Present 10 words- stop if 0 or 1 error • If >1 error, present next 15 words • If < 3 errors (out of 25), stop • If > 3 errors administer all 50 words Hurley & Sells, 2003
Expected PB max by Hearing Loss Dubno, J. et al.JSHR 1995
SPRINT: Speech Recognition Interpretation Chart Abnormally Low Word Recognition for HL PTA (dB HL) 2nd % Score 1st % Correct Score © by Linda M. Thibodeau, 1999
SPRINT: Speech Recognition Interpretation Chart 95% Confidence Limits for Word Recognition Score PTA (dB HL) 2nd % Score 1st % Correct Score © by Linda M. Thibodeau, 1999
Predicting WRS from the audiogram: The AI • The Articulation Index • Audibility Index • “Count the dot” audiogram • If word recognition is poorer than prediction: neural hearing loss or central disorder.
Counting Dots:(Mueller & Killion) • Each dot = 1% • # of dots below HL line on audiogram = % correct identification for speech.