Measurement - Quiz 6 - Speech Testing

Why do we do speech audiometry?

-Speech measures provide important info:

• regarding effects of reduced sensitivity on communication

• As a cross check of other test results (i.e. PTA/SRT)

• To assist with differential diagnosis

• For rehabilitative interventions and verification

  • need for and benefit from auditory prosthetics

• we also test “suprathreshold” speech processing abilities because they can provide

  • diagnostic information

  • information about communicative function

Why not just measure pure tone thresholds instead of speech audiometry?

PT thresholds aren’t always as predictive of speech understanding abilities

Types of speech tests

-Detection/awareness

• analogous to pure tone threshold detection

-Discrimination

• Same/different task

-recognition

• requires ability to label the stimulus

• Point to, write or repeat the stimulus

-Comprehension

• requires understanding of the stimulus meaning

Which types of speech tests can be part of the diagnostic test battery?

-detection/awareness

-recognition

Speech test materials

-single syllable phonemes

-multisyllabic nonsense syllable

-words

• monosyllabic words

• bisyllabic (spondee) words

-sentence materials

• nonsense sentences

• low context sentences

• high context sentences

• running speech

which speech test materials do we use?

-the most appropriate material depends on your test purpose

• diagnostic assessment?

• communication assessment?

• rehabilitative evaluation?

Performance intensity (PI) Functions for speech

-shows the change in speech understanding with changes in speech level

-note that for the same speech level, scores can vary dramatically depending on the test materials!

Factors affecting PI functions

-Task type

• detection versus recognition

• open versus closed set

  • number of items

-Context

• better understanding at lower levels with high context materials

-Familiarity

• better understanding at lower levels for familiar vs unfamiliar words

-Talker proficiency

• clear speech vs conversational

-Calibration

• the relationship between dB HL and dB SPL for recorded speech materials is variable!

Diagnostic Battery: speech tests

-Speech recognition threshold (SRT)

• also speech reception threshold (SRT) and spondee threshold (ST)

  • speech recognition is preferred

-Speech detection threshold (SDT)

• also speech awareness threshold (SAT)

  • SDT is preferred

-Word recognition score (WRS)

• also speech discrimination (SDS) or word discrimination score (WDS)

  • WRS is preferred

What is SRT?

-minimum hearing level (dB HL) for speech at which an individual can recognize approximately 50% of the speech materials

• hearing threshold for speech

-Test materials

• spondee (bi-syllabic) words

  • two syllable words with equal stress on each syllable.. airplane, baseball…

Why measure SRT?

-validate pure-tone thresholds

• SRT (in dB HL) should be very close to PTA (in dB HL)

  • very important cross check

-Quantify hearing loss for speech

• face validity

-provide a reference level for other speech tests

• ex: word recognition

-Estimate hearing loss in infants and young children (who cannot do pure-tone audiometry)

-assess/compare hearing aid benefit for speech

Why are PTA and SRT related?

-if you can’t hear the speech you can’t recognize and repeat the word…

• speech banana

What speech material should we use to measure SRT?

-the “best” speech material depends on the test purpose

• SRT purpose: find “threshold” for hearing speech

Why use spondees?

-for a threshold speech test, stimuli should:

• be familiar to the listener

• be phonetically dissimilar (not rhyming)

• contain a normal sampling of speech sounds

  • be representative of “normal” speech

• have equal difficulty across stimuli (homogeneity)

  • this results in a steeper, stable, psychometric function which helps make the test sensitive to variations in threshold

-original spondee words met these criteria

-Steep PI function for spondees suggests similar difficulty across test items (homogenous)

• lends itself to sensitive and reliable threshold estimates

• spondee scores increase from ~20-80% correct over a 6 dB range

How do we measure SRT?

-2 primary methods

1. ASHA method:

• currently used by veterans administration for compensation and pension (C&P) exams

2. Clinical method (martin & Dowdy, 1986)

• based on the modified Hughson-westlake procedure for measuring PT thresholds

• you’ll use this method in clinic

Instructions for SRT

-specific instructions can vary but must be understandable by the patient

• familiarize them with the words

  • details to follow

• explain the nature of the task

• tell them how to respond

-instructions should emphasize:

• only respond using words from the test

  • i.e. this is a closed set task

• respond even when faint, guessing is o.k.

-for example, after familiarization:

• “ok, I am going to say the same words again and I want you to repeat them back. But this time they are going to get very soft. repeat the words back as best as you can, and if you are not sure, take your best guess.”

Familiarization phase

-this phase ensures:

• the test vocab is familiar

• the patient can auditorily recognize each word

• the clinician can understand the patient’s responses

-to familiarize, give them a list of the test words

• present speech to both ears at a comfortable level

  • face to face/MLV is ok but no visual cues

• eliminate words that are difficult to identify during the familiarization phase

How many spondees to use for SRT?

-likely vary by clinician

-research says using recorded stimuli, a 15 word subset provides similar results to the full 36 word set

• ASHA guidelines provide 18 word half-lists and a 15 word short list

• but no work using MLV or for various talkers

ASHA Method SRT - Obtaining start level

-first, you have to find a starting level

• set the level 30-40 dB above presumed SRT

  • or 50 dB HL if no estimate is possible

• Present one word,

  • if correct, descend 10 dB

  • if incorrect, ascend 20 dB until a correct response is obtained,

    • then descend in 10 dB steps

• continue descending in 10 dB steps until…

  • one word is missed

• once one word is missed, present a second spondee at the same level

  • if they get that spondee correct, decrease level another 10 dB

  • Continue until two spondees are missed at the same level

• Once 2 spondees are missed at the same level, increase the level by 10 dB. this is your starting level

  • i.e. 10 dB above the level where 2 words were missed

ASHA method SRT - Threshold testing

-Descending procedure

• you can descend using either a 2 dB or 5 dB step size

-when using a 2 dB step size:

• present 2 words at the starting level and at each successive lower level

-for both step sizes, you must get five out of the first six words correct

• if not, increase start level by 4-10 dB and start again

-continue presenting 2 (or 5) words at each level and decrease in 2 (or 5) dB steps until stopping rules are met

-stopping rules: testing ends (stop decreasing levels) when:

• 5 of the last 6 words are incorrect (2 dB step size)

• 5 words presented at a single level are incorrect (5 dB step size)

-once you’ve met a stopping rule, you can calculate your SRT (threshold)

ASHA Method SRT - Calculating threshold

SRT = start level - # correct + CF (correction factor)

• 2 dB CF = 1 dB

• 5 dB CF = 2 dB

Clinical SRT Method

-Martin and Dowdy

• uses a modified Hughson-Westlake approach

  • use the same familiarization procedure as for the ASHA method

-Starting level - present a single word at a given level

• 30 dB, increase to 50 dB, then increase 10 dB steps until correct response

• after correct response, decrease 10 dB. this is your starting level

-Present a single word at each level

• down 10 dB; up 5 dB procedure

-Threshold: minimum of three correct responses, on ascending trials, at one level

• many audiologists use “2 out of 3” rule

Martin and Dowdy’s method pros and cons

-Pros:

• SRTs (recorded stimuli) and PTA - strongly correlated (r=.97)

• Test time ~2.5 minutes (similar or a bit faster than ASHA method)

-Cons:

• mean SRT was ~5 dB>PTA

• only ~50% of SRTs ± 5 dB of PTA

  • compared to ~78% for the current ASHA method

ASHA method vs Martin and Dowdy SRT method

-ASHA method provides better agreement between PTA and SRT

-less empirical work using MArtin and Dowdy method but…

• it appears to achieve the goal of validating PTA

• it is fast and familiar to clinicians

• it is being used routinely without evidence of systematic problems

-Either is fine!

Variations on the SRT theme

-Variations in response mode: whatever you pick it must be appropriate for the patient

• verbal or written repetition

• closed set picture pointing (use smaller set size)

-Presentation mode- MLV versus recorded

• flexibility versus standardization

• some work (and ASHA guidelines) suggest both methods are valid

Measuring spondee SRTs via MLV

-both spondaic syllables need to peak at 0 dB on the meter

-Adjust VU meter gain so you can talk naturally and peak the meter at 0 on both syllables

Reliability: MLV vs Recorded

-when experienced audiologists conduct testing both methods are reliable

• similar mean test-retest differences (both ~<1 dB)

• strong, positive test-retest correlations

  • but stronger correlation for recorded (.96) vs MLV (.89) presentation

• SRT-PTA correlations are similar, and high, with both methods (both ~.91-.92)

• Mean SRT-PTA differences are small and similar

  • but a little smaller for recorded, compared to MLV

    • SRT>PTA by ~1 dB for recorded and SRT<PTA by ~3 dB for MLV

Clinical guidelines for “agreement” between PTA and SRT

-Brandy suggests that aabsolute differences between SRT and PTA (2 or 3 frequency) of

• <7 dB- good agreement (no concerns)

• 7-12 dB- fair agreement (not great, but ok)

• >13 dB- poor agreement (look for cause of discrepancy)

-The VBWC follows this protocol

• others assume an SRT ± 10 dB of PTA is adequate

-based on both criteria the martin and dowdy method is acceptable

Speech detection threshold

-only obtain SDT if SRT not possible

• e.g. speech/language issues, cognitive impairment

-may use descending or ascending approach

• commonly use modified hughson-westlake method

-Materials: most speech materials work

• spondees, multiple words, running speech, single word repeated over and over

-SDT should approximate best PT threshold (250-4000 Hz)

• within 5-10 dB

• If difference >10 dB then question your results

-SDTs are generally better (lower) than SRT

When to mask SRT conceptually and how much

-mask if SRT of TE could be result of cross hearing, enough to effectively mask the NTE

-Initial masking level:

• SRT (NTE) + 10 dB safety factor

  • same as pure tone masking

-Maximum masker level:

• compare BC thresholds in TE to masker level in the NTE. You may be overmasking if:

  • Masker level-IA is >BC (TE) (PTA at .5, 1, and 2 kHz) or

  • Masker level - IA is > best BC (TE) (.5, 1, 2, or 4 kHz)

When to mask SRT Quantitatively

-Mask whenever the test ear and non-test ear SRTs differ by >IA

• assume speech IA = 45 dB (supra-aural) or 55 dB (inserts)

-Mask if:

• SRT (TE) - SRT (NTE) (or BC NTE PTA at .5, 1, 2, kHz) >45 dB

• Some rx: SRT (TE) - Best BC (NTE) (.5, 1, 2, or 4 kHz) >45 dB

General methods for assessing speech recognition

-there are two primary methods for quantifying speech understanding

• adaptive tasks- level/SNR varies across individuals

  • outcome: level/SNR needed for a fixed performance level

    • level/SNR needed to get 50% correct (or 71% correct, etc.)

    • we use this approach to determine our clinical SRT

• Fixed level/fixed SNR tasks- % correct varies

  • outcome: a “percent correct” score for a given level/SNR

-Word recognition testing in quiet is a “fixed level” task that is a part of the standard diagnostic test battery

-both methods have advantages and disadvantages

Slide 4