Measurement of Attention Bias - The Affective Lexical Decision Task

Various methods of operationalizing the expression of attention biases in depression have been created. Measures of attentional allocation popular in psychological information-processing research involve non-invasive procedures such as measuring reaction times or error rates in response to simple information processing tasks. The time it takes a research participant to react to an external stimulus is often suggested to correlate with the attentional resources allocated by the subject, to the task. Thus, for example, the time it takes subjects to name a stimulus (Powell & Helmsley, 1984), or to match one auditorily presented stimulus with another (Posner & Boies, 1971) is presumed to be proportional to the cognitive resources devoted to that task. Similarly, phenomena such as the interference of one attention-requiring stimulus with another in a selective attention task may be measured as a function of a participant's reaction time. For example in the classic Stroop interference paradigm (Stroop, 1935) subjects are shown cards on which the name of a color is printed in an ink color which may be congruous or incongruous with the color-name on the card. In general, people are slower to respond when the color-name on the card is different from the color in which the color-name was presented than when the color-name matches the ink color (Dyer, 1973). Theorists such as Posner and Snyder (1975) interpret this effect to suggest that attentional resources have been devoted, possibly unconsciously, to the word, thus, interfering with subjects' ability to name the ink color. This hypothesis amounts to a suggestion that on interference tasks reaction time will be shorter for stimuli to which the participant is not paying attention. For example, if a subject has an attention bias which prevents them from attending to positive words, and a positive word is printed on the card, the participant is not distracted from the color by the semantic content of the stimulus. Alternately, error rates on tasks such as the digit span task (Wechsler, 1974), in which subjects are asked to serially repeat as many digits as they can in a presented set, dichotic listening (Cherry, 1953) or dichoptic viewing (Neisser & Becklen, 1975) tasks are often suggested to correlate directly with the selective attentional resources devoted to the task (e.g., Lezak, 1983; Kahneman, 1973; Neisser, 1967; Neisser & Becklen, 1975).

One task which has frequently been used in the investigation of attentional biases in nonclinical as well as clinical populations is the lexical decision task. In this task a string of letters is presented to a research participant. The participant is then directed to decide whether the stimulus which has been displayed does or does not spell a word. Presumably, the time it takes the participant to make a lexical decision will be a function of the cognitive resources the participant devotes to ``attending'' to the stimulus, and the cognitive resources required for the participant to make the decision. Thus, for example, various studies have found that lexical decisions are faster for high frequency words than low frequency words (Chumbley & Balota, 1984, Gardner, Rothkopf, Lapan, & Lafferty, 1987), presumably because such words are more accessible to the participant and thus take fewer resources to match their visual image to some cognitive representation. Similarly, research has shown that lexical decisions are, in general, slower for stimuli which are degraded via blurring (Lorch, 1986), presumably because more cognitive resources are needed to identify such stimuli.

The lexical decision task is recognized as a particularly promising test of cognitive resources because reaction times associated with the task are hypothesized to be directly proportional to the cognitive load expended on the task (Coltheart, Davelaar, Janasson, & Besner, 1977; Rubenstein, Lewis, & Rubenstein, 1971). This is different from other traditional tests of cognitive resources which measure cognitive load more indirectly (e.g., interference tasks such as the Stroop task). Additionally, the lexical decision task provides both reaction time and error-rate measures such that techniques such as signal detection theory may be employed in its analysis. As shown by Neely (1977), reaction times to a lexical decision task have been shown to vary with variables assumed to mediate cognitive load, e.g., stimulus onset asynchrony and semantic relatedness to a previously presented prime, further suggesting that performance on the task may be mediated by semantic characteristics of stimuli.

While initial studies on the lexical decision task were conducted using affectively neutral stimuli (e.g., Chumbley & Balota, 1984; Lorch, 1986), recent interest in affective information processing, and information processing in the emotional disorders has prompted researchers to employ a version of the task using affectively valenced stimuli. The inclusion of words of varying affective content in a lexical decision task (e.g., positive, negative, and neutral words) is often hypothesized to allow explicit examination of the role attention to affect plays in the recognition of a stimulus, though delays in responding to words of a particular affective valence on the task could be construed as stemming from a number of cognitive mechanisms including attention to the affective aspects of the stimulus, or by association of the stimulus with other relevant cognitions which consume attentional resources.

In order to investigate attention biases in depression, a number of researchers have explored this type of a task with depressed and nondepressed people. In general, it is suggested that if depression influences attentional processes differentially with respect to the affective valence of a stimuli, reaction times to stimuli of differing affective valences will be different for depressed people. Differential reaction times may or may not be predicted to occur for nondepressed people based on different theories of affective information processing. Studies employing the task with depressed and nondepressed people have obtained varying results. The following sections will review the results of these studies, after which a discussion of their findings in light of theoretical models of attention in depression will be presented. A second methodology involving the identification of the affective content of stimuli, rather than a function of their semantic content will then be discussed as a method of obviating what is actually measured by the affective lexical decision task.