Study Population

For a study to be included in the review it had to meet the following requirements, patterned after Matt, Vazquez, & Campbell's (1992) meta-analysis of the relationship between mood and memory. Matt et al. describe rational for the exclusion of unpublished studies, cautions regarding the use of arbitrary labels such a ``good'' and ``bad'' or nosologic categories such as ``depressed''.

1.

The study had to be published between 1975 and 1995. Unpublished studies were excluded.

2.

It was required that studies clearly distinguish between the affective valence of stimulus materials, e.g., through reported pilot experiments explicitly devoted to obtaining norms for the affective valence of words, or through the use of a corpus such as Anderson (1968) for the likeableness of personality trait adjectives. In the studies reviewed, valence was commonly labeled ``negative'' or ``bad'', ``positive'', or ``good''.

3.

Studies had to include as an outcome measure the response latencies to the different valences of stimuli. Some studies reported mean response latencies only to correctly identified stimuli, while others reported mean latencies to all stimuli. Due to the typically very low error rates reported in these studies, these reporting techniques were equated.

4.

Studies had to meet minimum reporting requirements including the mean response latency for each group in each condition, standard deviations or equivalent information (e.g., confidence intervals) which would allow calculation of standard deviations around those means, the number of subjects in each condition, and the number of words of each affective valence used. Letters were sent to the contact author of all studies meeting other criteria but not minimum reporting requirements. Studies for which the author responded by sending the necessary data were included in the meta-analysis.

5.

Studies examining performance on the affective lexical decision task as a function of variables other than those labeled depressed, sad, nondepressed, neutral, elated were also excluded (e.g., anxiety, schizophrenia).

6.

Only studies which conducted the lexical decision task in the absence of a semantic prime condition (displaying a word for a brief period before displaying the target word) are reported due to the hypothesized interaction of priming with affective recognition (Conway & Bekerian, 1987; Matthews & Southall, 1991). Three studies exclusively employing a priming methodology were excluded (Hartley, Ireland, Arnold & Spencer, 1991; Hill & Kemp-Wheeler, 1989b; Kemp-Wheeler & Hill, 1992) and data from three studies in which a priming condition was used in conjunction with a nonpriming condition were excluded (Bradley et al., 1994; Conway & Bekerian, 1987; Matthews & Southall, 1991).²

All available studies reporting the use of an affective lexical decision task with nondepressed, depressed, or induced depressed subjects were collected using the PsycINFO (American Psychological Association, 1992) database and a rigorous literature review. Twenty studies met al.l inclusion requirements. Within the twenty studies, forty relevant contrasts were identified each yielding one effect estimate for differential reaction times to words of differing affective valences.