Why an Attention Test Is Harder With Tones Than Letters

What the research found

“Press the button as soon as you hear the high tone. Do nothing when you hear the low one.” An instruction of roughly that form opens one of the most frequently administered attention tests in child assessment, and it appears to ask for nothing beyond patience.

The continuous performance task presents a long stream of stimuli with a target buried in it at intervals, for eleven minutes. Missed targets are recorded as lapses of attention and mistaken presses as impulsivity. Response speed is taken as a measure of processing time.

Hettie Roebuck, Claudia Freigang and Johanna Barry, of the MRC Institute of Hearing Research in Nottingham, gave that task to the same participants three times, altering only the signal. Their report appeared in the Journal of Speech, Language, and Hearing Research in 2016. Twenty adults and twenty children aged 8 to 12 completed all three versions in counterbalanced order, each screened beforehand for normal hearing, normal nonverbal reasoning and no history of attention or language difficulty.

One version used the letters X and O on a screen, a second the same letters spoken aloud, the third two pure tones, a low C and a high G, identified to the child only as “high” and “low”. Everything else was held constant: 72 targets among 324 stimuli, each presented for a fifth of a second, two seconds apart.

The children missed 9.6 percent of targets on average when the target was a tone, against 2.8 percent for the spoken letter and 2.0 percent for the letter on screen, and both tone comparisons survived statistical correction. False alarms ran in the same order, 6.5 percent for tones against 1.4 percent for speech, and responses to tones were the slowest, averaging 820 milliseconds against 568 for the visual letter.

Adults made too few errors for that comparison to be informative, but their reaction times ran in the same direction, 415 milliseconds for the letter on screen and 660 for the tone. Presentation order changed nothing.

Roebuck and colleagues attribute the tone penalty to demands the other versions avoid. “High” and “low” name a relationship between the two tones rather than a thing, so the child must hold the comparison in mind and verify it before each response, and with no word or shape to attach the sound to, that comparison rests on short-term memory alone. Children who missed more tone targets had shorter forward digit spans, a correlation that did not survive correction and was confounded with age, so the memory account remains the authors’ reading rather than their result. The size of the tone penalty itself was not in question.

What it means for you and your child

Not one of these children had any history that suggested difficulty attending, and changing what they were asked to detect still moved their omission rate by a factor of nearly five. Whatever a continuous performance task registers, it registers the cost of recognising the signal along with the cost of staying with the task.

Recognising a signal covers more than hearing it. The child has to encode the sound and settle which of the two it was, inside two seconds, while the next stimulus approaches. A letter arrives already labelled; a tone has to be labelled on the spot, every time, for the length of the test. That labelling work is what separated the versions, in children whose attention was never in question. Five of the twenty needed the practice run repeated before testing could begin, all of them on the tone condition.

The test manufacturers have long behaved as though this were true. Norms for the visual version of the Tests of Variables of Attention, the commercial task these researchers modelled their parameters on, begin at age four; norms for the auditory version begin a year later. Error rates in this study fell with age in both auditory conditions and not in the visual one, which the authors read as auditory processing maturing on a longer schedule than visual. An auditory percentile is therefore already scored against a later starting line.

What the study cannot do is tell any family that a given score was inflated. These were twenty typically developing children in one laboratory, compared against themselves rather than against a clinical group, and no child with ADHD was tested. On what the instrument is sensitive to, the evidence here is plain.

That sensitivity matters when a report quotes percentiles from a computerised attention test, because the stimulus is the detail such reports omit most often. Two children can receive the same percentile from tasks that differed in difficulty for reasons having nothing to do with attention.

The precise question to put to whoever administered the test is which version the child sat: letters on a screen, spoken sounds, or tones. Where it was tones, the useful follow-up is whether the visual version was given as well, and what separated the two. Ask how much of that gap belongs to the signal.