Task Analysis and Chaining: Teaching Skills That Have More Than One Step

Many of the skills that matter most to independence are chains — sequences of discrete responses that have to occur in order. Handwashing, dressing, making a sandwich, using a vending machine, a morning routine. Teaching them requires two things: breaking the skill into steps, and choosing how to teach the steps.

Task analysis: breaking it down

A task analysis is the written sequence of steps that make up the skill. Getting it right is more than half the work.

Three ways to build one:

• Observe a competent performer. Watch someone who does the skill fluently and write down what they do. Fast and usually accurate.
• Consult an expert or an existing published task analysis. Useful for complex or safety-critical skills.
• Perform it yourself and record each step. Surprisingly effective at surfacing steps that competent performers do automatically and forget to mention.

Getting the grain size right

The most common task analysis error is steps that are too large for the learner. Wash hands is not a step. Turn on cold tap is.

Grain size is individual. A learner with good motor skills may handle apply soap as one step; another may need press pump once, move hands under dispenser, and rub palms together separately.

The test: if a learner consistently fails a step, it is probably two steps. If they master every step immediately, the analysis is finer than it needed to be and you are spending sessions on precision you did not require.

Validating it

Before teaching, run the task analysis with someone who has the skill and confirm the sequence produces the outcome. Task analyzes written at a desk routinely omit a step, and a missing step becomes an unexplained failure point that gets misattributed to the learner.

The three chaining procedures

Forward chaining

Teach step one to criterion. The instructor completes the remaining steps. Then teach steps one and two, and so on.

Advantage: the skill is taught in its natural order, which suits sequences where early steps genuinely gate later ones.

Disadvantage: the learner does not complete the chain — and therefore does not contact the natural reinforcer at the end — until late in teaching. For a learner whose motivation depends on the outcome, that is a long wait.

Backward chaining

The instructor completes every step except the last. The learner performs the final step and immediately contacts the natural reinforcer. Once mastered, the learner performs the last two steps, and so on backwards.

Advantage: the learner completes the chain and contacts the terminal reinforcer from the very first session. This is a significant motivational advantage and the main reason backward chaining is chosen.

Disadvantage: it can feel counterintuitive to staff and families, and it requires the instructor to perform most of the chain early on, which takes time.

Total task presentation

The learner attempts every step in every session, with prompting on the steps they cannot yet do. Prompts fade step by step as competence develops.

Advantage: the learner experiences the whole sequence every time, which supports fluency and keeps the chain intact. Often the most efficient of the three for learners who can already perform some steps.

Disadvantage: more demanding per session, and it requires the instructor to manage prompting across many steps simultaneously.

Choosing between them

• Total task when the learner can already perform several steps, the chain is short to moderate, and they tolerate the full sequence. This is the most common sensible default.
• Backward chaining when the terminal reinforcer is powerful and motivation is the limiting factor, or when the chain is long and the learner would otherwise wait a long time to complete it.
• Forward chaining when early steps must be fluent before later ones are safe or possible, or when the natural order carries strong stimulus control.

Data on a chain

Chain data is step-level, and this is where it goes wrong. Recording handwashing: 60% is nearly useless — it does not say which steps failed, and the answer changes what you do next.

What to record per session:

• Each step scored independently — independent, prompted (with prompt level), or incorrect.
• The pattern across steps, not just the total. A learner failing steps 4 and 9 has a different problem from one failing steps 8 through 11.

The step-level pattern is diagnostic. Consistent failure at one step usually means the grain size is wrong there, or that step needs its own teaching outside the chain. Failure at everything after a certain point often means the chain is too long for the learner's current stamina.

Practically, this means the data system has to hold a task analysis as a structured sequence with per-step scoring, rather than treating the whole chain as one target with a percentage. Programs recorded as a single percentage lose exactly the information a BCBA needs at review.

Common problems

• Steps too large. The most frequent cause of a stalled chaining program.
• An invalidated task analysis with a missing step, producing a failure point everyone blames on the learner.
• Prompting the whole chain indefinitely because no per-step fading criteria were written.
• Teaching in one setting only. Chains are especially prone to setting-specific stimulus control — a learner who washes hands in the clinic bathroom and not at home has learned the clinic bathroom.
• Ignoring the natural cue. The chain should eventually be occasioned by the natural discriminative stimulus — dirty hands, arriving home — not by an instructor saying wash your hands. If the verbal instruction is never faded, you have taught a prompted routine rather than an independent skill.

The summary

Write the task analysis at a grain size that matches the learner, validate it by running it, and choose the chaining procedure from the learner's motivation and current repertoire — total task as a sensible default, backward chaining when motivation is the constraint, forward chaining when order gates safety. Score every step separately, read the pattern rather than the percentage, program across settings, and fade the instructional cue so the natural one takes over.