A review of studies examining the nature of selection-based and topography-based verbal behavior.

Roche et al. (1997) read every paper they could find on two ways we talk. One way uses mouth moves, like saying "cookie." The other way uses finger moves, like tapping a picture of a cookie.

They wrote a story about what the papers showed. They did not run new kids or new trials. They just sorted the old facts.

The mouth-move kind almost always comes first and feels easier for most kids. Once those mouth words are solid, the finger-tap kind is simpler to add.

So, teach talking first, then add pictures or devices later.

Petursdottir et al. (2023) looked again at the same split and said, "Stop counting mouth vs. finger. Count how big the picture array is, how alike the icons look, and how fast a kid can build a word." Their idea updates the 1997 view: pick AAC tools by those real specs, not by the old label.

Lowenkron (1991) helps explain why the 1997 pattern happens. He showed that finger-tap words only spread to new pictures if the child can already say the word out loud with a partner. That joint-control skill is the bridge the 1997 review talks about.

The papers do not fight. Roche et al. (1997) tell us the order; Petursdottir et al. (2023) tell us the specs to watch once we pick that order.

Start early learners with spoken mand and tact drills. Keep pictures and SGDs ready, but do not rush. Once the child can echo and say new words with you, bring in the device. When you do, check icon size, array load, and visual clutter first—not just whether it is "selection-based." You will save weeks of re-training.