In defence of iconic memory

According to a model of visual information-processing which originated with Sperling (1960), and which currently enjoys wide acceptance, the contents of brief alphanumeric displays are initially held in a high-capacity fast-decay visual-information store (“iconic memory”). Some of these items are subsequently transferred to a more durable form of storage; the remaining non-transferred items are lost. Observers can select which items are to be transferred on the basis of physical characteristics of the items (such as colour, location, size, shape or brightness).

This model has recently been attacked by Holding (1970, 1971, 1972, 1973), sometimes by claiming that iconic memory does not exist, and at other times by claiming that transfer from iconic memory cannot be selectively controlled by the observer. We argue in this paper that Holding's criticisms are incorrect, and that, even if they were correct, the experiment we report would not be open to objections he has raised concerning previous studies of iconic memory. Despite this, evidence fully supporting the orthodox model was obtained, and we therefore conclude that this model remains tenable.     

Non-spatial delayed alternation by the pigeon

Pigeons were trained on a non-spatial delayed alternation task in which the color not pecked on the previous trial was correct. When varying delays were interposed between trials, alternation accuracy decreased as a function of delay, but remained greater than chance with a 45-sec delay. Successful alternation on the longer delays was accomplished without behavioral mediation of the delay intervals. Also, during initial testing when a position cue was available in addition to the color cue after incorrect trials, alternation accuracy was greater after a preceding incorrect trial than after a correct trial. When the position cue was removed, no differences occurred as a function of the outcome of the preceding trial.     

Discrimination of checkerboard patterns by rats

Fifteen rats were trained to discriminate between a regular checkerboard pattern and a similar pattern containing a “mistake.” The animals subsequently transferred to many other pairs of regular and irregular patterns differing widely from the originals. The results suggest that rats can induce and store highly abstract rules describing input patterns.     

A stochastic model for free recall

A statistical model for verbal learning is presented and tested against experimental data. The model describes a Markov process with a realizable absorbing state, allowing complete learning on some finite trial as well as imperfect retention prior to this trial.This work was carried out while the author was at Lincoln Laboratory, Massachusetts Institute of Technology.Operated with support from the U. S. Army, Navy, and Air Force.