The objects of attention: causes and targets

tel.: +46 (0)46 222 0897
ingar.brinck@fil.lu.se
http://213.80.36.53/temp/filosofen3/staff/person.asp?id=17&lang=eng

Abstract

The objects of attention can be located anywhere along the causal link from the source of stimuli to the final output of the vision system. As causes they attract and control attention, as products they constitute targets of analysis and explicit comments. Stimulus-driven indexing creates pointers that support fast and frugal cognition.

Hurford suggests that the objects of attention should be understood as indexed, arbitrary objects identified by their location in a mental, spatial map. Objects of attention are available to the subject without categorisation or encoding of their properties or locations.

I do not agree with Hurford’s characterisation of indexed objects as arbitrary and identified by their location in a mental map. First, indexing is not really arbitrary, but stimulus-driven. Not any object will be indexed, but only those that are salient enough to impinge on the subject. Indexing is caused by some property of the object, although that property will not be encoded (Pylyshyn, 1999; Pylyshyn, 2000). Furthermore, at the moment of indexing, the objects are distinguishable as visual patterns or clusters in the visual field.

Finally, the spatial map is not mental, but the scene in the real world forms a local map that contains the indexed objects. The scene itself does not have to be memorized. Indexed objects serve as pointers that allow the subject to access and revisit locations in a distal environment without engaging attention. Thus indexed objects support fast and frugal cognition, which exploits information in the environment (Brooks, 1991; Hutchins, 1995).

It is difficult to see how indexed objects could be objects of attention. We can think of objects of attention as either causes, attracting attention, or effects, that is, products, of focal attention. Objects of attention conceived of as effects constitute enduring targets to which attention can be maintained. They are analysed and, if kept in short-term memory long enough to reach consciousness, they can be intentionally commented on by the subject (Weiskrantz, 1997).

In order to allow for sustained attention, a target object must at least make it possible for the subject to track the object by its properties. The subject encodes the properties in short-term memory. Objects of attention cannot be discriminated by mere location, since the identification of locations relies on a previous segmentation of space (Driver & Baylis, 1998). Minimally, target objects are constituted by segmented regions which form unities also when in motion. They have some spatiotemporal consistency.

I do not specifically address the question whether objects of attention are objects, features, or locations, but take it that attention is directed to objects (O’Craven, Downing & Kanwisher, 1999; Yantis, 1998). The objects of attention can in principle be located anywhere along the causal link from the source of the stimulus to the final output of the vision system. Which properties will be ascribed to the objects of attention depends on the level of analysis (Eilan, 1998). The properties reflect the various cognitive roles of the objects of attention.

Objects of attention can be introduced on at least three levels of analysis. On an initial, preattentive level, they constitute the input to the early vision system, and are best thought of as causes. This level processes segmented objects for focal attention and subsequent analysis. On a computational and attentive level, further on in the early vision system, objects of attention constitute targets that are processed in the dorsal and ventral paths. The dorsal and ventral paths may construct different and incompatible representations from stimuli from the same source, without the subject’s noticing it (Goodale et al., 1994).

On a psychological, or phenomenological, level, which receives output from the early vision system, the objects of attention will be multimodal, three-dimensional percepts. Percepts occur on a personal level and are directly available for the organism as a whole, as opposed to being processed subpersonally. The subject may become consciously aware of them and choose to comment on them (Weiskrantz, 1997). Comments are voluntary and intentional and can be communicated through behaviour or language. A comment will be cognitively penetrable if sensitive in a rational, or semantically coherent, way to the organism’s goals and beliefs (Pylyshyn, 1999).

Hurford furthermore suggests that the objects that subsequently are indexed attract attention, treating them as causes of focal attention. He claims that certain "natural attention-drawing properties" of the objects attract attention. These properties concern the biological needs of the subject and are highly encapsulated. In contrast to the percepts that are arrived at after an analysis in the ventral stream, these properties are not accessible to the subject on a personal level. Information about them is exchanged only between the subsystems of vision.

I do not see the need to introduce "natural attention-drawing properties" to account for attention attraction. I agree that whatever it is that attracts attention, it must be of interest to the subject. An "abstract" object cannot be so, simply because it is propertyless. Attention is attracted by objects that have an informational, and not merely causal, impact on the subject (Brinck, 2001). They are at odds with what the subject is expecting on the basis of previous experience. But the impact is not necessarily related to biological needs, or to positive or negative values. Except for cases when an anticipated object attracts attention, the object will only receive a value to the subject once it has been detected (Eimer, Nattkemper, Schröger & Prinz, 1996).

I submit that so-called goal-driven attention works top-down, in anticipation of some well-defined item. The subject is searching for a particular object, and the attention is geared to react when it appears (Ballard, Hayhoe, Pook & Rao, 1997; Yantis, 1998). The subject's needs and desires determine the aim of the search. The salient feature that serves to indicate the appearance of the object is likewise selected before the search begins.

Stimulus-driven attention, on the other hand, works bottom-up. Attention is attracted by sudden and unexpected changes in the subject’s immediate environment (Freyd 1987). Expectancy relates to familiarity. The change must introduce a new and somehow anomalous object or feature in the visual field in order to draw the attention of the subject (Yantis, 1998; Yantis & Johnson, 1990). It seems as well that the saliency of the object will increase if the object is behaviorally relevant according to the needs or drives of the subject (Gottlieb & Goldberg, 1998).

To sum up, indexed objects as described by Hurford can only serve as pointers. If conceived of as objects with properties (albeit not encoded), they also take on the role of a cause that controls the subject by inducing her to index them. However, indexed objects can never be targets of attention. They are mere place-holders.

References