What representations might underlie infant numerical knowledge?

The sensitivity of young infants to number raises the question of the nature of their numerical representations. We contrast two broad classes of models that have been proposed in the literature: Integer-symbol models and Object-file models. Four infant-addition experiments (1 + 1 = 2 or 1) were conducted. Experiments 1 and 2 showed that the timing of the placement of the screen relative to the objects determined 8-month-olds' success (object-first condition) and failure (screen-first condition) in such addition tasks. Experiment 3 established that 10-month-old infants succeed in the single-screen, screen-first, paradigm. Experiment 4 examined whether the failure in the screen-first condition was due to an imprecise representation of object location. Separate screens enabled X-month-olds to succeed in a screen-first paradigm. We argue that this pattern of success and failure favors the Object-file model over Integer-symbol models.