One way to perform such a test would be to have research participants respond directly to the affective valence of a stimulus (e.g., to indicate whether a stimulus is negative, positive, or neutral). It is hypothesized that depressed people will, in general, respond faster to the negative words than to neutral or positive words on this task, as presumably, they have overlearned associations with stimuli possessing a negative affective valence. This phenomenon should not be present for people who have not overlearned negative associations (i.e., nondepressed people). These results are hypothesized to hold, even when performance on a lexical decision task does not show such facilitation. Thus, the ``valence-identification'' task is expected to be more sensitive to overall negative attention biases in depression than the lexical decision task. As with the lexical decision task, the amount of ruminative effort in which subjects engage during the task may be manipulated by varying the SD over a short range. If subjects are prone to ruminate on pathologically negative cognitions in the absense of a stimulus, a short SD condition might reveal facilitation on negative, or delays on positive words more than a condition with a longer SD.
Valence-identification or ``affective decision'' tasks have been used sparingly in the literature. Mathews and Milroy (1994) used an affective decision task to assess sensitivity to threat words with anxious people, and found relatively little difference between affective decision times to positive and negative words while finding facilitation on threat words with this group. Such a result would be predicted under the integrative framework if anxiety involves the overlearning of threatening stimuli, and not positive or negative stimuli. Hill and Kemp-Wheeler (1989a) also describe an affective decision task in which subjects are asked to categorize words as threatening or nonthreatening. They find strong effects for anxious research participants, and relatively small differences in affective decision time for depressed participants. This finding is also expected as there is no immediate reason to believe that depressed people would have overlearned anxiety provoking words.5 A literature search revealed no experiments using an affective decision task with a depressed population.
Weaver and McNeill (1992) used a variant of the task in which subjects were required to appraise the affective valence of entire sentences. They studied differences in reaction time latencies for nondepressed subjects and subjects in whom a depression had been induced. The authors state that a mood by valence interaction with response time latency was not statistically significant, but do not report any statistic which would allow a calculation of effect size for this measure. Since the authors were analyzing a three way interaction (Mood x Gender x Sentence affect) and since their sample sizes were very small (N=24 for one version of the task and N=32 for another) and since the authors used only 12 happy and 12 sad sentences, it is not clear whether their apparently null results were due to a true absence of mood congruence effects or power too low on some dimension to detect a true difference between the groups of interest.
Possibly the first use of a valence-identification task was by Derryberry (1988) who employed a valence-identification methodology in an attempt to isolate mechanisms which might influence emotional information processing including general arousal, spreading activation, and selective attention. Derryberry created a task in which a negative, neutral, or positive warning signal was presented, followed by a positive, negative, or neutral word to which subjects had to press a key indicating the word's valence, followed by a positive, negative, or neutral feedback signal. Derryberry reports finding a significant Valence by Warning by Feedback interaction effect with both reaction time and error rate, making it difficult to interpret how subjects would have responded were there to have been no warning or feedback tones. Simple effects analyses suggest that words were categorized faster based on an incentive to perform which is congruous with their affective tone. Were this incentive to act as a momentary mood induction, the results would be explainable by the present theory. Yet, Derryberry also finds that positive words are categorized fastest after feedback involving loss, while negative words are categorized fastest after feedback prompting joy. Given that the actual effects of incentives and feedback on people's moods are difficult to predict, the results are difficult to integrate completely with the present theory. Still, results do suggest that the task is sensitive to affect-related variables, and thus examining the task in depressed vs. nondepressed individuals may yield interpretable results.
Importantly, the same theories which predicted differential valence-mediated performance on the lexical decision task may also be reinterpreted to make predictions regarding the affective decision task. For example, adherents to Bower's (1981) network theory would predict that the reciprocal activation from the stimulus to the affective node of its valence would allow quicker recognition if that valence was already activated. This would happen regardless of the semantic determination which is made for the word. Were a person paying attention primarily to the negative affective valence of the stimulus, as suggested above, their reaction time would be decreased to all negative words on the valence-identification task. Their reaction time would be increased to negative words on the lexical decision task. Similarly, Nolan-Hoeksema's idea of rumination on negative symptoms of depression would provide for quicker access to the valence of any negative stimulus than the valence of a positive stimulus, though the semantic content of the stimulus might not be so accessible. Were rumination a trait, as suggested by Nolen-Hoeksema and Morrow (1993) this finding would be true for people who are currently depressed, and also for formerly depressed people. Interestingly, Stip et al.'s (1994) theory of affective stress suggests a different outcome. Were depressed people overcome by stress at the negativity of a stimulus, presumably they would not be able to respond quickly either to the valence-identification or the lexical decision task.
The affective valence identification task also allows discrimination between the affective interference theory and a similar but slightly different explanation of lexical decision task performance. Niedenthal and Setterlund (1994) suggest that emotion congruence experiments with induced depression using lexical decision tasks generally find null results because stimuli are not semantically close enough to the induced emotion to take advantage of facilitation. It might be assumed that small differences between reaction time latencies to positive and negative words on the lexical decision task by depressed people occur because mental representations of the negative words used in the task are not well-connected to the mental representations for negativity or sadness. Were such a phenomenon to occur, one would predict relatively little facilitation on the affective valence-identification task for the same stimuli. If, instead, the mental representation of the associated affective valence was strongly activated and competition were to occur between multiple activated semantic representations in the lexical decision task, strong facilitation would be predicted to occur on the valence-identification task for depressed people on negative words.
Another theory, developed by Mathews and Milroy (1994) to explain affective information processing in anxiety, can also be used to predict delays on the affective lexical decision and valence identification tasks task. They suggest that anxious people may well be predisposed to attend to threatening stimuli, but, at the same time, intentionally avoid such stimuli once they are identified. In this way, they suggest, people who are anxious will be delayed in tasks requiring semantic elaboration of a word (e.g., the lexical decision task), but, should be facilitated in simply recognizing the word's valence. In fact, Mathews and Milroy (1994) do find such a facilitation for anxious subjects. Potentially the same explanation might be proposed for depression if the predicted results are obtained. Yet, such a theory would seem to predict that valence identification happens before processes involved in making a lexical decision can operate. As reaction times on the one lexical decision task conducted with anxious people (Hill & Kemp-Wheeler, 1989a) were, in each case, faster than those reported by Mathews and Milroy (1994), some doubt may be cast on the use of their theory for explaining information processing biases in depression, though the presentation styles, and other variables related to the experiments differed a great deal. Similarly, affective decision times, as reported by Mathews and Milroy (1994) are much slower (means ranging from 1214ms to 1540ms for different conditions) than those for analogous lexical decision task experiments (means ranging from 556 to 1041ms for different experiments in different conditions).6 It will thus be important to investigate relative speed of reaction times on the valence-identification and lexical decision tasks with regard to negative and positive words in depressed and nondepressed individuals.