Methodology

Our Methodology

Each feature is a published behavioral procedure, and the parameters below are the ones the app runs. Prompt fading uses 4 levels; SAFMEDS timings last 60 s; the chart spans 140 days × 6 log cycles; FSRS-ABA learning steps are 1 min → 10 min → review.

1

Programmed Instruction with Prompt Fading

Programmed instruction comes from B. F. Skinner's teaching-machine work: small steps, an active response to every frame, and immediate feedback (Skinner, 1958, 1968). Errorless-learning research showed that strong prompts, faded gradually, keep errors low while stimulus control transfers from the prompt to the question itself (Terrace, 1963; Touchette & Howard, 1984).

The goal is transfer of stimulus control: by the end, the question alone should evoke the answer. Each frame carries four prompt levels, from most to least help:

L0 · Full
A sentence that contains the answer. You read it and type the term.
L1 · Partial
The first letter, blanks for the rest and the letter count.
L2 · Intraverbal
A related statement with no letters. You recall the term from the words alone.
L3 · Independent
No prompt. The question alone.
Start
Every session starts at L3, so you first get a chance to answer without help.
Time delay
Each question shows alone for 2 s before its prompt and the answer box appear.
Fade criterion
2 consecutive correct responses at a level move you up one level.
Error correction
Any error returns you to L0, so the next response is very likely correct.
Feedback
Immediate: right or wrong, the target term, a one-line explanation and its source.
Programmed-instruction frame RBT-A.1 at level “L0 · Full prompt”: the stem about a stopwatch started at the onset of a tantrum, and the full prompt that contains the answer, duration.
Figure 1. After a miss, the next frame comes up at L0 · Full prompt. The answer is inside the prompt, so the next response is likely correct.
2

Standard Celeration Chart

Ogden Lindsley built the Standard Celeration Chart (SCC) in the 1960s to standardize how rate of response is displayed. Its first standard version, the Standard Behavior Chart, dates from December 1967. Because every chart uses the same axes, any two charts can be compared at a glance.

Vertical
Count per minute on a logarithmic (multiply) scale, 0.001 to 1,000 per minute: six cycles. Equal ratios are equal distances, so ×2 from 5 to 10 looks the same as ×2 from 50 to 100.
Horizontal
140 successive calendar days (20 weeks). Days you did not time stay visible as gaps.
Celeration
Change in rate per week, written as a multiplier. ×2 means the rate doubles each week; ÷2 means it halves. Because the grid is fixed at the SCS standard 8 × 5.28 (20 weeks × 6 cycles), a ×2 line always rises at 33.5°, the “34° rule.”
Record floor
1 ÷ counting time: 1 per minute for a 1-minute timing. A zero is drawn just below it.
Plotted
Your best SAFMEDS timing each day: corrects as a dot, that same timing's errors as an ×.
Fit
One line per phase (needs 3+ charted days). Default: least squares through log rate. Switchable in the app to the quarter-intersect method (White & Haring, 1980): split the days in half and draw the line through each half's median day and median rate. The two agree on steady data; quarter-intersect resists outliers.

celeration = e7b, where b = slope of ln(rate) per daySo a fitted slope of 0.099 ln-units per day gives e0.693 = ×2.0 per week.

Digital and hand-drawn chart styles are both available in the app.

The app’s celeration chart after a demo sprint: daily dots for corrects and crosses for errors on six log cycles, aim line at 40, today’s dot at 49 correct per minute, celeration lines labelled ×1.7 and ×1.6 per week.
Figure 2. “Your recent sprints” in the app, straight after a sprint. Earlier dots are the account’s own timings; the last dot (49/min) is the scripted demo sprint.
3

Fluency Building (SAFMEDS)

Accuracy alone is not enough. Carl Binder (1996) defined fluency as accurate performance at speed, and traced the findings linking fluent performance to retention (it lasts without re-teaching), endurance (it holds up over longer work periods) and application (it transfers to new situations).

SAFMEDS means “Say All Fast a Minute Each Day Shuffled.” Ogden Lindsley introduced it in the late 1970s as an improvement on flashcards.

Timing
60 seconds, fixed. The deck is shuffled every time.
Reading pause
The clock pauses while each new card is read (about 252 ms per word, at least 600 ms), so long stems don't penalize you.
Grace period
If you are mid-word at the bell, you get 3 s to finish. That answer gets feedback but is not counted.
Scoring
Corrects and errors are counted separately and converted to per-minute rates.
Aim
40 or more correct per minute, with errors near zero.
Chart
Your best timing of the day becomes one dot, plotted with that timing's own errors. Corrects and errors are never mixed across sprints. Every sprint is still saved in the sprint log.

Classic SAFMEDS uses spoken answers to physical cards. Here you type or tap, so your rates are not directly comparable to published aims for spoken SAFMEDS. Compare yourself with yourself, using the chart.

SAFMEDS sprint mid-run: 53 seconds left, the card “A child reaches for a toy without requesting; the RBT withholds the toy and waits for a verbal request, then immediately delivers it; this is ____ teaching.”, the answer “incidental” being typed, counters 5 correct, 1 error, 6 total.
Figure 3. A SAFMEDS sprint in progress: 53 s left, 5 correct, 1 error.
4

FSRS-ABA Adaptive Scheduling

Spaced repetition schedules each review near the point of forgetting. Ebbinghaus (1885) described the forgetting curve; Pavlik & Anderson (2008) computed practice schedules from a memory model; FSRS (Ye et al., 2022) tracks each card's stability and difficulty. Our scheduler, FSRS-ABA, borrows that per-card model and grades each answer the way a behavior analyst would: by accuracy and latency.

Grade
Error = again. Correct in under 4 s = easy, 4–9 s = good, over 9 s = hard (stems longer than 10 words get 0.3 s more per word).
Latency
Timed from when the answer box opens, not from when you press submit.
Learning steps
1 min → 10 min → graduate to spaced review. The first spaced interval is 3 days, stretched to 4.5 for an easy answer or cut to 2.4 for a hard one.
Interval growth
Faster correct answers grow the interval more. Growth scales with how much of the last interval actually passed, so reviewing early earns little or no spacing credit. A correct answer never shortens an interval.
Error recovery
A miss on a review card cuts stability to a fifth, schedules a 1-day relearning step and returns prompting to L0.
Difficulty
Each card's difficulty (1–10) rises with errors and slow answers; harder cards get shorter intervals.
Mastered
In review, memory stability ≥ 21 days, and no more than 2 lapses.
RBT dashboard: 127 frames due now, 0 mastered (stability ≥ 21 days), 53 reviewed of 129, 50 due earlier today, plus the exam-date plan grid with checked-off, missed and rest days.
Figure 4. What the scheduler produces: frames due now, mastered at stability ≥ 21 days, reviewed, and the day-by-day plan to the exam date.
5

Answer Matching & Error Detection

A typed answer is correct only if it contains every key word of an accepted target. “Negative reinforcement” is never accepted for “positive reinforcement.”

Rule
All key words present, in any order.
Typos
One edit (a wrong, missing, extra or swapped letter) for words of 4+ letters, or the same first 4 letters and a length within ±2 for words of 5+ letters.
Never merged
Look-alike real words such as cost/most, last/least, taken/token, cause/pause.
Ignored
A leading “a”, “an” or “the”, capital letters and curly quotes.
Errors
Anything else counts as an error and is plotted on the chart.

Target: positive reinforcement

  • ✓
    reinforcement, positiveOrder doesn't matter.
  • ✓
    postive reinforcmentOne-letter slips pass.
  • ✗
    negative reinforcementA key word is wrong.
  • ✗
    reinforcementA key word is missing.
6

BACB Task List Alignment

Every task in the current BACB test content outlines has three programmed-instruction frames: definition, example and application.

Exam TrackEditionTasksContent Areas / DomainsFrames
BCBA6th Edition1049 content areas (A–I)312
BCaBA6th Edition909 domain areas270
RBT3rd Edition436 domain areas (A–F)129

Content status (v2026.10.19): BCBA and BCaBA mapped to the 6th Edition outlines (September 2024 update); RBT to the 3rd Edition (effective January 1, 2026). 711 frames in total (312 BCBA, 270 BCaBA, 129 RBT), 3 per task. AI-assisted and machine-validated: one blank per frame, no answer in the stem, a valid APA citation, the correct task code and no copied BACB text. Not yet independently BCBA-reviewed; this study aid does not guarantee exam outcomes.

Report a concern: email hello@behaviorceleration.click with the track and task code; verified corrections ship in the next content release. Implementation questions: request a consultation.