Resource allocation and the attentional demands of letter encoding

The idea that familiar events can be encoded automatically has gained general acceptance in cognitive psychology since Posner and Boies (1971) first reported that reaction times to a secondary probe were not interfered with by letter encoding. More recently, Ogden, Martin, and Paap (1981) used a more valid control for estimating baseline probe performance and found secondary task interference, suggesting that letter encoding does require attentional resources. The present series of experiments began with the aim of evaluating Ogden et al's evidence against automaticity when the first letter was not terminated after a brief exposure, as was done in their study. In the first set of experiments we found evidence of encoding interference when the interval between the two letters was varied (50 to 1,000 msec), but this interference disappeared when there was a constant 1,000-msec interval between the letters. On the basis of these findings, we hypothesized that changes in the primary task (e.g., the exposure duration of the first letter or the interval between the two letters) may influence the momentary allocation of resources between the primary and secondary tasks. More specifically, we hypothesized that any momentary reduction in the resources demanded by the primary tasks results in a reallocation of resources to the secondary task, which in turn reduces the sensitivity of the secondary task to the demands of the primary task, that is, probe performance is moved into the data-limited region of processing (Norman & Bobrow, 1975). This idea was tested by reducing resource allocation to the probe task at the time of encoding by reducing the expectancy (i.e., the probability) of probes in the temporal proximity of the first letter. The results showed that this manipulation produced a large and significant increase in encoding interference. Moreover, when the intensity of the tone (probe) was decreased from 70 to 60 dB, the magnitude of encoding interference was further increased. In regard to the specific issue of automaticity, the findings suggest that encoding familiar events does require resources, which will result in secondary task interference given that the secondary task is in the resource-limited region of processing. More important, the findings suggest that the magnitude of secondary task interference is dependent on within-trial changes in resource allocation between the primary and secondary tasks. This possibility has general implications for dual-task methodology and the measurement of attentional demands.