Logan (1992)
Are automatic and preattentive processing identical? Can they be distinguished theoretically and empirically?
Preattentive processes provide information necessary for attentional processing. For example, a simple feature must be detected before a stimulus can, on the basis of that feature, be selected for further processing. Preattentive processes and automatic processes are often equated. For example, in early filter theories of attention, elementary physical features (e.g., color) are said to be processed without attention. Unfortunately, it is nearly impossible to rule out the possibility that unattended stimuli are in fact unattended. As Logan points out, small effects of unattended stimuli may be due either to preattentive processing or to occasional lapses of attention. In short, "Postattentive automatic processing may be mistaken for preattentive, attention-free processing" (p. 319).
According to the modal view of automaticity, automatic processing is processing that occurs in the absence of attention. By definition, preattentive processing is therefore automatic. However, automatic processes are not necessarily preattentive. Automatic processes can follow or bypass attentional processing. According to Logan, a problem with the modal view is that automaticity is defined in terms of what it is not. Furthermore, the development of automaticity is explained only in terms of the gradual withdrawl of attention, with no mechanism to explain the withdrawl.
Automaticity may also be viewed as a memory phenomenon. According to this perspective, novices solve problems using a general algorithm, whereas experts simply retrieve solutions. Stated another way, novices attend to the steps of the algorithm, whereas experts attend to the solutions stored in memory. In both cases, automatic processes are dependent on attention. What differs is the level of task organization attended to.
Logan fleshes out the view a postattentive view of automaticity in his instance theory. Three major assumptions are the basis of this theory. First, attention to a stimulus is sufficient for memory encoding. What is attended to will be encoded. Second, attention to a stimulus is sufficient for retrieval of information about that stimulus from memory. That is, all that is known about a stimulus will be retrieved from memory when it is attended to. Finally, each encounter with a stimulus is represented as an instance, which is encoded, stored and retrieved separately. Obligatory encoding and retrieval strengthen the memory trace for familiar experiences, and obligatory retrieval makes past experiences available for present problems.
An alternative account of automaticity (e.g., Schneider & Shiffrin) proposes that automaticity is preattentive and reflects the ability of a stimulus to attract attention automatically. The acquisition of this ability is called priority learning. Logan describes two experiments to test the priority learning hypothesis. In Experiment 1, subjects performed a visual search task in which the stimulus was presented alone, or in which the stimulus was presented with unpronounceable nonwords. Experiment 2 was the same except that the divided attention condition was substituted for a selective attention condition in which the target word was colored word. Reduction of the difference between RTs in the focused attention conditions and the divided and focused conditions (i.e., focused RT – divided/selective RT) would provide evidence for automatic attention attraction.
Near asymptotic slopes in the focused, divided, and selective conditions provided evidence for automatization in all three conditions. Nevertheless, the difference between the focused and the divided and selective RTs persisted. As Logan explains, "This suggets a dissociation between automaticity and preattentive processing, in that one can be obtained without the other" (p. 329). This finding is inconsistent with the view that automatic processes become preattentive, and that the two are identical. Curve fitting suggested that additional practice would not have narrowed the gap. Similar results were obtained in an additional experiment contrasting selective and divided attention RTs, although extrapolation suggested convergence of asymptotes. Nevertheless, at least on the first 16 trials, there was clear evidence for dissociation between automatization and priority learning.
The results are inconsistent with the view that automatic processes are preattentive. For example, even though divided attention performance reached asymptote, indicating automaticity, the difference between focused attention performance and divided attention performance persisted with practice. This pattern of results suggests a dissociation between preattentive and automatic processes.
Logan argues that the common feature of all automatic processes is that they depend on attention, are postattentive. Preattentive automaticity is, by contrast, automatic in a different sense, namely, in that it is independent of attention. Thus, from Logan’s perspective, "The conventional view that automatic processing is independent of attention would lead to the unhappy conclusion that much of automaticity in not automatic" (p. 336). An important question remains, however: How can automatic processing be defined without reference to independence from attention?
Logan suggests that preattentive processing—processing in the absence of attention—and postattentive automatic processing can be distinguished by the types of computations they perform. On the one hand, preattentive processing are supported by local parallel processing. Computations that required spatial indexing or reference to information in disparate visual regions are not possible. On the other hand, postattentive automatic processing is predicated on direct-access memory retrieval.
Logan concludes with a consideration of terminological issues concerning preattentive and postattentive automatic processing. He suggests that it should be possible to capture the essential aspect of preattentive processing—independence from attention—without invoking the term automaticity. By contrast, he argues that postattentive automaticity is "a phenomenon in itself" which can justifiably be called automaticty.
Preattentive vs. automatic processes
Preattentive processes and automatic processes are often equated. And, in fact, they behave similarly in many ways. For example, both can be carried out concurrently with another task with no performance decrement in either task. There are a number of important differences, however. Treisman et al. (1992) argues that the origin of independence for these processes is different. Automatic processes are highly specific, and are mediated through memory traces consisting of conjunctions of featural information. This conclusion is supported by findings which, for example, show that conjunctive search, but not feature search, is affected by minor changes in the stimulus display. Preattentive processes are by contrast governed by populations of feature detectors.
Logan, on the other hand, argues that automatic processes are not, in fact, attention independent. Indeed, he suggests that attention is required for the initiation of an automatic process. He provides evidence for this claim by showing that, while performance on both a focused attention task and a more attention demanding divided attention task both reach asymptote (indicating automaticity), the focused attention advantage persists throughout training. This is interpreted as evidence against Schneider and Shiffrin’s (1977) view that automatic processes become preattentive via priority learning—that is, through the development of automatic attention attraction.
In conclusion, then, it is safe to say that preattentive processes are independent of attention—both Treisman et al. (1992) and Logan (1992) would agree with this assertion. Where Treisman et al. and Logan differ is in the role of attention in automatic processes. Logan argues that automatic processes are postattentive and inextricably bound to and dependent on attentional processes, while Treisman et al. seem to believe that automatic processes can, in fact, achieve independence from attention.