Naming Pictures Among Other Pictures: Does a Related Image Slow You Down or Speed You Up?

Back in 1989, Glaser and Glaser ran a study with a tidy result. Ask someone to name a picture, and if a second picture sitting next to it belongs to the same semantic category, the naming takes longer than it would if the neighboring picture were unrelated. A dog beside a cat, say, versus a dog beside a hammer. The slowdown looked real and reliable.

Most models of how people produce speech have a ready explanation for this. Somewhere at the level where words get chosen, both pictures supposedly wake up their corresponding names, and those names then fight each other for the one selection slot. Two competitors from the same category interfere more than two unrelated ones, so the target name comes out slower.

We think that story deserves a second look. The assumption doing all the heavy lifting is that a context picture automatically activates its own name, and a body of newer work on picture naming and word translation points the other way. Context pictures, it seems, do not reliably light up their names on their own. If that assumption is shaky, the interference result built on top of it is shaky too.

So we ran two experiments. The first was essentially a replay of Glaser and Glaser's setup, and it reproduced their interference effect: related context pictures did slow naming down. But the data also hinted at a different culprit. Participants sometimes appeared to lock onto the wrong picture, selecting the context item instead of the target, and that mix-up could produce the slowdown without any name-versus-name competition.

The second experiment tackled that confusion directly. We changed the task so that picking out the correct target picture became easier, removing much of the opportunity for the context item to hijack selection. Under those conditions the interference vanished. What showed up instead was a modest effect in the opposite direction: naming the target was slightly faster when its neighbor was semantically related.

A small facilitation effect is a very different animal from a robust interference effect, and it does not sit comfortably with models that lean on lexical competition to explain semantic context. We work through what the reversal means for those accounts of language production.