Matching to sample

One trial has several defined parts

A matching-to-sample trial includes:

1. a sample stimulus
2. a comparison array
3. a response window and selection rule
4. a correct relation for that sample
5. a programmed consequence or probe condition

If the sample is a photo of a bus, selecting the identical photo may be correct. If the sample is the printed word “bus,” selecting a bus photo may depend on an arbitrary learned relation.

Identity and arbitrary matching differ

Identity matching pairs physically identical or defined identical stimuli. Arbitrary matching relates stimuli that do not match physically, such as an object and its name.

Both require conditional control. The bus picture cannot be the right answer on every trial; it is correct only when the relevant sample is present. Otherwise, a learner may select one favored comparison without observing the sample.

MTS is also a measurement abbreviation

In ABA data collection, MTS can mean momentary time sampling, a discontinuous measurement system. Spell out the term in plans, graphs, and records. “80% MTS” could otherwise refer to trial accuracy or sampled intervals.

Intended relations can be replaced by shortcuts

A person may select by position, border color, teacher gaze, item size, or a predictable sequence. High accuracy can therefore reflect unintended stimulus control.

Balance comparison positions and sample order. Vary irrelevant features. Use neutral presentation and independent probes. Fisher and colleagues embedded identity matching within a prompting hierarchy for two autistic children, illustrating one way to assess observing behavior in a narrow sample.

Grow and colleagues compared receptive-labeling teaching methods and reported detailed procedural-integrity measures. The study shows why prompts, positions, and observing responses need direct measurement.

A fictional matching record

Omar practices 12 independent trials. Six are identity matches between object photos, and six are arbitrary printed-word-to-photo relations. Comparison position is balanced and no feedback follows probe selections.

Omar responds correctly on 10 of 12 trials. He scores 6 of 6 on identity matches and 4 of 6 on arbitrary relations. The aggregate is 83.3%, but the relation-specific results show a meaningful difference. The session does not prove stimulus equivalence, generalization, or control by the intended feature.

Stimulus equivalence requires additional tests

Sidman and Tailby distinguished conditional discrimination and matching-to-sample procedures from the additional relations used to demonstrate equivalence. Directly taught A-to-B and B-to-C performance is baseline evidence. Reflexivity, symmetry, and transitivity require untrained tests under defined conditions.

Use precise language: “matched trained relations at 90%” reports what occurred; “formed an equivalence class” makes a broader claim that needs the complete evidence.

Accessibility affects valid performance

Check visual contrast, auditory access, tactile features, motor effort, array spacing, language, device use, and response latency. Provide AAC and other communication throughout. A missed response may reflect inaccessible materials rather than a weak relation.

Eye contact is never required. A person can observe a sample through any reliable sensory and motor route. Avoid unnecessary trial volume when fatigue or discomfort makes the data less valid.

The BACB BCBA Test Content Outline, 6th edition includes conditional discriminations, matching-to-sample, prompting, and stimulus equivalence in examination content. It is not a protocol.

Report enough detail to reproduce the claim

Record correct independent selections divided by eligible trials for each sample and relation. Also report prompts, errors by position, invalid trials, response latency, comparison number, correction procedures, and presentation integrity.

Keep teaching, correction, independent probe, maintenance, and generalization trials separate. If a correction repeats the sample, say whether it begins a new trial. Document novel stimuli before testing them so exposure after an error does not remain labeled untrained.

Array size changes chance and scanning demands

Two-comparison trials provide a different chance level and visual search demand from four-comparison trials. Report array size for each relation and avoid pooling them without explanation. An increase in errors after adding comparisons may reflect scanning, crowding, motor effort, or an underlearned relation.

Assess whether the person observes the sample before selecting. An accessible observing response can be looking, touching, listening, repeating, signing, or another reliable action. It should support contact with the sample without imposing eye contact or unnecessary compliance.

When mastery is important, include independent trials across balanced positions and enough presentations of every sample. A single high total can hide that one sample never controls the intended selection.