Speech in Quiet

Speech in Quiet covers the speech-audiometry measures you take in a quiet booth, with no competing noise. In Theta it isn't a single test but a group of three you can mix and match in one element:

• SRT — Speech Recognition Threshold (the ~50% spondee point)
• WRS — Word Recognition Score (PB-max, the percent of words repeated correctly)
• Sound Sensitivity — the loudness measures: SAT (Speech Awareness Threshold), MCL (Most Comfortable Level), and UCL (Uncomfortable Loudness Level)

You choose which of these the learner performs, the values are generated from the patient's hearing, and the learner discovers them at a simulated speech audiometer that looks and behaves like the pure-tone audiometer — two channels, ear and transducer selectors, and a masking channel.

Like every element, speech in quiet follows the same lifecycle — Design → Settings → Testing → Scoring → Report. This page walks the lifecycle once, calling out what's specific to each of the three tests as it goes.

The three tests

All three sit on one Performance-Intensity Function (PIF) — a curve of percent correct against presentation level. SAT anchors the bottom (0%), SRT the 50% point, WRS the peak at MCL, and UCL marks the ceiling. You'll see this curve in the designer, in the saved-runs view during submission, and on the report.

Design

Tests in this lesson

Every new speech test opens on its Design tab, and the first thing on it is Tests in this lesson: a tile for each of SAT, SRT, WRS, MCL and UCL. Tick the ones the learner performs. At least one stays ticked — tapping the last one says so instead of unticking it. A test saved before this rule with nothing ticked opens with SRT and WRS ticked, and new attempts run those; attempts that had already started keep what they had (see the end of this page). The element is named by what it runs (SRT, or SRT · WRS), and a new element starts with the tests your element preset names.

Unticking a test never throws anything away: its values stay designed, so ticking it again brings them back, and another speech test on the same audiogram can still run it.

Ticking a test brings back the element's Review scoring badge, because the test's row on the Scoring & Report tab is new. Unticking a test doesn't bring the badge back.

Below the tiles the tab has two parts, kept apart: the values, in a small table per ear, and the controls that shape the patient. A test this element doesn't run keeps its values, because they belong to the shared speech patient, but shows them lighter and read-only, and its controls are hidden. To change one, tick the test here, or open the speech test that runs it.

A line under the tiles names the audiogram the test belongs to and any other speech tests that share it; each name opens that element. Tests are named as the learner's tabs name them, so two that run the same tests read WRS 1 and WRS 2, here and on the audiogram's strip. The values and the controls are the shared speech patient, so a change here shows at once in every other speech test on the same audiogram.

Values

One small table per ear, Right (red) and Left (blue), side by side on a tablet or wider and one above the other on a phone. Each row holds one value and nothing else:

• On a fixed case, the values of the tests this element runs are boxes you type over. On a template every value is read-only, because it is one draw.
• The values of tests this element doesn't run are lighter and read-only. One line under the tables says why: they aren't tested in this element, and they come from the shared patient.

Beside the tables on a wide screen, and under them on a phone, is the live performance-intensity chart. Editing a value reshapes the curve at once. The landmarks of tests this element doesn't run are faded, not removed, because they still shape the curve.

All five measures are always designed even if you only run or score some of them, so the curve is fully anchored.

Controls

Under the values, in a section of their own, are the controls that shape the patient, grouped by test and set per ear. Only the tests this element runs show their controls. The others are hidden, not greyed.

The section's header holds Reset N typed values on a fixed case with values typed by hand (see below), and Shuffle on a template.

Which tests this one runs is chosen at the top of this tab. Its word lists and how it presents live in its Settings tab. The WRS list size, the category cut-offs, the speech earphone and masking belong to the audiogram and are set in the audiogram's Settings. How each test is scored lives in the Scoring & Report tab.

Template vs. fixed

This is inherited from the linked audiogram, not set here. A tag beside Values says which it is:

• Template (violet): the audiogram is a template, so each learner gets a freshly generated patient. Speech values are calculated and read-only. You shape the patient with the controls (SRT–PTA agreement, WRS category, rollover) rather than typing exact numbers. Every speech test on the audiogram gets the same patient in each attempt.
• Fixed (emerald): the audiogram is fixed, so every learner sees the values shown here. They start out calculated from the audiogram, and you can type over the values of the tests this element runs. The chart updates live as you type.

Shuffle

On a template case, the Shuffle button in the controls draws new thresholds for the audiogram and a new speech patient for every speech test linked to it. It is the same button as the audiogram's own Shuffle, so you can preview another patient from either element. A fixed case has no Shuffle on the speech tab: you type over the values instead.

The audiogram's own Shuffle, and changing its profile so its thresholds are drawn again, also draw a new speech patient. Nothing else draws one: opening a lesson, switching between elements or saving never changes a speech value. Editing the audiogram's thresholds moves the calculated values with them (see below), and the controls move the values they describe.

On a fixed case, the audiogram's Shuffle and a new profile ask first when anything has been typed by hand, on the audiogram or in the speech tables, because the new draw replaces it.

Typed values (fixed cases)

A value you type becomes part of the design. The controls name the values typed by hand (Typed by hand: Right SRT, Left UCL). Leaving a box empty goes back to the value the audiogram gives, and Reset N typed values clears them all. Both cover only the tests this element runs: another speech test's typed values keep their design. Values you haven't typed keep following the audiogram: edit a threshold and the calculated SRT, SAT, MCL, UCL and WRS move with it, while the typed ones stay put.

Typed values that make a strange case get a warning, not a block:

• an SRT far from the PTA for the chosen SRT–PTA agreement
• a SAT above SRT − 5
• an MCL at or above the UCL, or at or below the SRT
• a presentation level at or above the UCL, or less than 20 dB above the SRT
• a word score far above what the audiogram predicts

The warnings also cover only the tests this element runs. A value typed on a test it doesn't run is warned about in the speech test that runs it, where you can change it. The two presentation-level warnings are the exception: they show in the element that runs word recognition, whichever test the UCL, MCL or SRT was typed on, because that element's presentation method sets the level.

No response (SRT)

A profound loss routinely has a speech awareness threshold and no reception threshold at all — the patient detects speech but never repeats a spondee, at any level. On a fixed case, tick No response for that ear in the SRT controls to design that. The SRT then reads NR in the table.

The level beside the box stays required. It is the highest level the case was run to, usually the audiometer ceiling (120 dB HL). Ticking the box starts it at the SRT's current level, so type the level the case was run to over it. It is still the signal level the masking window is built from, which matters here more than anywhere — the other ear will cross-hear those loud spondees and repeat them, so the learner cannot call no response until they mask. The right answer is the learner pressing No response in the console; a typed threshold is wrong. Leaving that level box empty unticks the box and goes back to the audiogram's SRT.

Leaving the SRT blank is not the same thing. A blank value means "not designed": the console refuses to present the ear at all and scoring skips it.

An ear with no reception threshold has no 50% point, so it has no performance-intensity curve and no word recognition either. The designer says so in a blue notice, the WRS category control says so for that ear, the console will not present words to that ear, and its WRS rows are not scored. Whenever the audiogram puts the SRT above the audiometer's ceiling, the ear is no response on its own: on a template for every draw that lands there, and on a fixed case with the box ticked for you, where it stays ticked.

The loudness sweep needs all three landmarks

SAT, MCL, and UCL drive one simulated patient reaction to continuous speech, so all three need values even if the lesson only asks the learner to record one of them. The designer flags the missing ones — without them the Sound Sensitivity tab has nothing to react against and will not play.

Settings

The Settings tab is the case-wide configuration that isn't a designed value.

It opens with one line naming the tests this element runs — Tests in this lesson: SRT · WRS — with a Change link back to the picker on the Design tab. Scoring is independent and set in the Scoring & Report tab: a test can be part of the lesson but not scored.

The sections follow the ticked tests: WRS configuration and WRS classification ranges show only while WRS is ticked, and SRT configuration only while SRT is. The audiogram, general settings, masking and voices always show. In an element preset every section shows, and the same tiles at the top — Tests a new Speech in Quiet element runs — pick the tests a new element starts with.

Audiogram

The audiogram this test is calculated from. With one audiogram in the lesson it's simply named. With several, pick one here. Moving a test to another audiogram gives it that audiogram's speech patient. There is no "not linked" option: every speech test belongs to an audiogram.

WRS configuration

WRS classification ranges

The minimum percent for each category. A score at or above a threshold falls into that category; Very poor is anything below Poor. Like the word count, these are shown here and set on the audiogram, so every test on it names a score the same way. Defaults:

SRT configuration

Pick the spondee list the learner uses for SRT, and whether each spondee is introduced with the carrier phrase ("Say the word…"). It is separate from the WRS carrier setting and off by default: spondees are usually presented on their own.

General settings

Voices

Who the learner hears. Every line is a recorded voice (the same on every computer and browser), not the browser's own speech.

Voices are presentation only: changing them never changes the answer key.

When the lesson's speech elements give the learner patients who may sound different — different accents, two different patient voices, a fixed voice beside Random, or two Randoms with different examiners — a notice under the voices names them (Speech in Noise's WIN counts too; its other tests show the patient's replies as text). The same voice in two accents is two different patients: each accent is its own recording. It's allowed: one patient per lesson is the ideal, not a rule.

The accepted masking window

When a test needs masking, the answer key carries a window of masker levels rather than a single number, derived the way Theta 3 did it:

• Minimum = speech level − IA + the non-test ear's largest air-bone gap + safety pad (the least masker that covers the crossover with a margin — an air-conducted masker loses the non-test ear's gap before it reaches that cochlea).
• Maximum = the test ear's best bone threshold + IA (above this the masker crosses back and starts masking the test ear). For WRS the crossback only costs words once it pulls the test ear's best-frequency sensation level below 50 dB (conversational level) — at PB-max on a sloping loss the good low frequencies sit 65–90 dB SL, and a masker 10–20 dB into a band that far above threshold changes nothing the words depend on — so the WRS maximum is IA + max(best bone, presentation level − 50). On a sloping loss that is a 20 dB band for supra-aural phones (60 for inserts). SRT, a threshold test, keeps the bone cap.

The speech level is the SRT for SRT masking and the presentation level for WRS. "Best bone" and the air-bone gap are read over the bone frequencies the learner is required to test in that ear — the audiogram's required bone set, the same frequencies Theta 3 compared. A normal 250 Hz bone threshold carries almost no word intelligibility, so it only caps the masker when 250 Hz is part of the set the ear reasons over. When no bone frequency is plotted for an ear — the ear One-ear BC leaves untested, or an ear with every bone frequency unticked — the band falls back to that ear's stored required-bone grid, then to its required air frequencies (a learner who never tests bone in an ear compares speech against what they have), and finally to the whole axis; the generated bone values at those frequencies, 250 Hz included if it is in the set, still feed best bone and the air-bone gap.

For WRS the window is re-derived at the level the learner actually presented the graded run at: any level inside the accepted presentation range is judged against the physics at that level (both edges move with the presentation level), so the answer key and the simulated patient always agree.

The Design tab's Masking controls show the window as e.g. 35–95 dB EM; the report shows the same range as the expected value and grades the learner's saved masker level against it. If the minimum is above the maximum the case is a masking dilemma — no single level satisfies the physics. The learner's level is then graded as mask enough, and no more: from the minimum up to one plateau width (the Plateau setting, 10 dB by default) above it, since every extra dB is more crossback onto the test ear. The report shows that band as e.g. 60–70 (dilemma), and the Masking controls as 60–70 dB EM (dilemma); their heading says a dilemma even while they're folded away. The designer also shows a blue "Masking dilemma in this case" notice for any such ear, with the size of the unavoidable crossback, what it does to the simulated patient's WRS at the minimum adequate masker, the graded band, and whether insert earphones would open the window. It is a notice, not an error: the case still plays and scores. With the WRS audibility rule no sensorineural shape produces it; expect it on a conductive non-test ear whose air-bone gap the window cannot absorb, and for SRT on a conductive test ear.

Testing

The learner works at a speech audiometer with Channel 1 (speech) and Channel 2 (masking noise), each with its own level and ear. Both channels wear the transducer the case was designed for — the picker shows it but is locked, because the answer key is built on that transducer's interaural attenuation. A prominent header shows the live presentation level and the saved values for both ears, and a response strip shows what the patient "said" (neutrally: a right answer and a wrong one look the same). Channel 1's stimulus is the word list for SRT and WRS. Each channel has a power button: Channel 1 starts on, and while it's off nothing is presented; Channel 2 starts off. Channel 2 also has a noise picker: Speech noise (speech-shaped, the clinical default) or White noise. Each channel has a volume slider (above its level) for what the learner hears; it never changes the level the patient gets.

The test is heard as well as shown: the examiner's word plays in the test ear (with "Say the word…" when the carrier phrase is on), the patient's reply follows, and the masker plays in its ear while Channel 2 is on. Sound Sensitivity reads the Rainbow Passage aloud, then the patient says how loud it was. Keyboard shortcuts mirror the audiometer: arrows change the speech level, Shift + ↑/↓ (or Page Up/Down) the masking level, M toggles masking, R/L/E switch ears, Space presents, and Enter is the context action (the full list is in Keyboard shortcuts). A Show audiogram toggle puts the learner's own audiogram beside the speech panel, so a masking decision can be read off the non-test ear's bone thresholds without leaving the tab. Until the learner has tested the audiogram, it says so instead of showing an empty chart.

Presenting at or above the ear's designed UCL gets a complaint ("Ouch, that's too loud") instead of a word, for SRT and WRS alike — as in Theta 3.

The enabled tests appear as tabs. A green check on a tab marks a saved result for the current ear. With a single tab (an SRT-only test, say) the tab bar is left out: there is nothing to switch between. In the lesson's own tab bar, a speech test is named by the tests it runs, e.g. SRT or SRT · WRS — just the tests, as a Speech in Noise element is named by its test. A test that runs all five is just Speech in Quiet.

SRT

A grid of spondee words is shown. The learner sets the speech level, then clicks a word to present it; the patient either repeats it or gives no response. With variability off the patient repeats every spondee at the designed SRT and none 5 dB below it, so bracketing lands on the designed value; with variability on, the same near-threshold probability table as the audiometer applies (off entirely while the masker plays). Pressing Save SRT records the current level as the threshold, or No response marks that the patient never identified the words — which is the correct answer on an ear designed as no response, and wrong on one that has a threshold.

Every presentation is logged to a presentation history (level, word, what the patient heard, correct/incorrect), and an optional per-level tally view helps the learner home in on the 50% point. The carrier phrase is shown (and said) only when the SRT carrier setting is on.

When SRT–PTA agreement is scored, an SRT–PTA agreement panel under the SRT asks the learner to call each ear Agree or Disagree. Their own PTA appears beside the call once they've entered one on the audiogram (when the audiogram asks for a PTA), or from the start when the audiogram is given to them filled in. (Given only its unmasked thresholds, the PTA is still theirs to work out, so it appears once they enter it.) If the SRT comes before the audiogram in the lesson, the learner comes back to make the call after finding the PTA. The call can be changed, or cleared by pressing it again, until the lesson is submitted.

The call is only asked when the learner can see the audiogram — a hidden audiogram they never test gives them no PTA to judge by — and whether it is asked is settled when the attempt starts, so a later change to the scoring never grades a call the learner wasn't shown. Where it isn't asked, the SRT is checked against the PTA automatically, as before.

WRS

The learner presents each word in turn from the active list (10/25/50/100 words). The WRS panel plays out the exchange: the examiner's line is spoken word by word ("Say the word boat." when the carrier phrase is on, or just the word), then after a short pause comes the patient's reply. That reply is the word, one of the designed foils, a "didn't catch it", or no response, and it looks the same whether it's right or wrong, so the learner has to judge it. A word missed outright is the foil half the time, a "didn't catch it" about a third of the time, and silence otherwise, unless the run is scored by phoneme, where the foil is always given so the phonemes that survived can be marked. Changing the presentation level or the masker part-way through a list raises a notice — the remaining words are answered at the new settings, so the run mixes two performance levels — and the learner marks:

• By word — ✓ Correct / ✗ Incorrect for the whole word.
• By phoneme — chips for each phoneme; click the ones the patient missed.

A running score (e.g. 18/25 (72%)) tracks the current run. A list left part-way is saved with the attempt: switching to another tab or element, or leaving the lesson, and coming back picks it up at the same word, level and masker, with the same responses. Submitting saves it first too, as far as it got, but only saved runs are scored. Save run stores that run at the current presentation level. Because rollover lives at high levels, the learner can re-test the list at several levels and save each run — a saved-runs view lists them, with a Runs / PI function switch that swaps the table for a plot of the runs on a PI function so a high-level drop is visible. The table shows first; the choice is remembered in that browser. When rollover scoring is on, a Rollover — positive / negative control asks the learner to make the clinical call after testing above MCL.

WRS responses are masking-aware: a masker that crosses back to the test ear reduces the effective level, and the non-test ear can pick up the signal through cross-hearing (reduced by masking at that ear), exactly as in a real masking dilemma.

Sound Sensitivity (SAT / MCL / UCL)

SAT, MCL, and UCL share one Sound Sensitivity tab built around a continuous speech passage (the Rainbow Passage). The learner sets the level and presses Play speech; the words light up and the patient gives a loudness reaction — (no response), Very soft, A little soft, Just right, Comfortably loud, Getting too loud, or Too loud! — which is fully deterministic from the designed SAT/MCL/UCL. The learner sweeps the level and saves each landmark with its own Save SAT / Save MCL / Save UCL button.

Scoring

Open the Scoring & Report tab. It starts with the same one-line Tests in this lesson summary as Settings. The first panel lists every test in the lesson with a tri-state control — Scored (counts and on the report), Report only (on the report, not counted), or Off (performed but neither scored nor reported). See Scoring overview for how this tri-state works across elements. Only tests set to Scored get the detailed options below.

All measures score by value tolerance (in dB, or % for WRS) — the submitted value is correct if it's within tolerance of the designed value. Beyond that:

SRT scoring

• Value tolerance (dB). On a No response design the value is graded on the call instead: the learner is correct only if they pressed No response.
• Score masking decision — was masking used when it was needed (and not when it wasn't)? Noise routed into the ear under test does not count as masking.
• Also score the masker level — did the saved masker level fall inside the accepted window (see above)? Turn this off to grade the decision only, for a lesson that teaches when to mask but not yet the SRT level rule, or a programme that teaches a different formula.
• Score SRT–PTA agreement — the learner's own Agree / Disagree call for each ear (see Testing). The right answer comes from the designed SRT and PTA — the PTA of the audiogram's PTA type: Agree when they are within the Agree within tolerance (dB), else Disagree. So a learner who mis-measures their SRT can still make the right call, and no call at all is marked wrong. A no-response ear is not graded, because it has no threshold to compare. Where the call isn't asked (see Testing), and on attempts started before it existed, the learner's SRT is checked against the PTA automatically instead: right when its agreement with the PTA matches the case's own.

WRS scoring

• Score tolerance (%).

• Score presentation level — was PB-max tested at an appropriate level?

• Score classification — does the percent land in the right category?

• Score masking decision and mask level classification. With the level scored, Masker level is correct when… picks the rule:

  • Inside the accepted band (default) — anywhere in the derived window above, judged at the level the graded run was presented at: covers the crossover, keeps the test ear at or above 50 dB SL.
  • Exactly PL − IA + largest ABG + safety pad — the formula the masking manual teaches. The pad is the same safety pad from the masking parameters (10 dB by default; edited on the audiogram's Speech IA block), so the formula answer is the bottom edge of the accepted band above.
  • Exactly PL − 20 dB — Studebaker's rule.

  Every rule is judged from the level the learner actually presented the graded run at. The two formula rules are graded exactly: any louder masker is over-masking, any softer is under-masking. Pick one when your programme teaches that formula and expects it to the dB; the band is for "did the masker do its job". A masker routed into the test ear never counts as masking: the masking-decision row grades it as "not masked" and no level row is emitted.

• Score rollover — the learner's positive/absent call. Turning this on is what surfaces the rollover control in the submission.

SAT / MCL / UCL scoring

Each has a single value tolerance (dB).

Report

The report assembles every test marked Scored or Report only.

• PIF chart — a combined Performance-Intensity Function for both ears. The designed curve is drawn per ear (red/blue) with filled markers at the included anchor points (SAT at 0%, SRT at 50%, WRS at MCL), and dashed verticals for MCL and UCL when those are included. The learner's submitted values are overlaid as small × marks, including each saved WRS run — so a rollover pattern reads at a glance.
• Data table — for each included measure, the designed value, the submitted value, and whether it was scored correct, alongside the WRS classification it fell into. Tests set to Off never appear; Report only tests are shown but don't affect the score.

Every attempt keeps the tests it started with. Changing the tests (or their scoring) later never changes what an attempt was asked or how it is graded, and editing the scoring of a test an attempt didn't run (or ran with scoring Off) doesn't mark that attempt's results as settings changed. Attempt replay draws the same tests: an SRT-only attempt replays its SRT, not an empty word recognition sheet beside it, and each loudness landmark the attempt asked for is listed even before it was saved.

That holds for an attempt that started with no tests ticked, too — which only a test saved before every speech test had to run one can have. The learner saw No speech tests are turned on for this lesson, so the attempt still shows that, scores nothing, keeps its plain Speech in Quiet name and replays no tests. Its grade never changes, even though the designer now shows that test with SRT and WRS ticked.