Verbal Overshadowing

Verbal overshadowing (VO) refers to situations in which describing a nonverbal experience, such as the appearance of a face, impairs subsequent recognition. In the original demonstration of VO, participants viewed a video of a robbery with a salient perpetrator and were later asked to recognize the perpetrator in a photographic lineup, including seven similar distractors. After viewing the video, half the participants spent 5 minutes writing a description of the robber, while the other half performed an unrelated control task. Remarkably, participants who verbalized the face were significantly less likely to correctly identify the target face among others in the lineup.

Since its original demonstration, considerable research has explored the generality and boundary conditions of VO. VO has been found to generalize to a large variety of other visual experiences, including colors, abstract figures, photographs of mushrooms, and maps, as well as other sensory moralities, including audition (e.g., music and voices) and taste (wines).

In contrast, it is not observed for more verbalizable experiences such as the contents of spoken statements. The boundary conditions associated with VO involve both situation and individual difference variables. An important individual difference variable is expertise, with verbalization impairing performance when perceptual expertise is high (e.g., when individuals have been perceptually trained with a stimulus) and verbal expertise is low (e.g., among individuals with high verbal ability or those trained to describe a stimulus).

A variety of situation variables have been found to mediate VO during the encoding, postencoding, and retrieval phases. With respect to encoding, verbalization is observed with faces of persons from one’s own race but not with faces of those of other races (demonstrating a role of perceptual expertise). With respect to postencoding, extensive verbalization produces greater VO than more modest descriptions. With respect to retrieval, VO is more likely to be observed when the distractors are highly similar than when they are dissimilar (and thus more verbally distinguishable). VO is more likely to be observed for faces presented upright than inverted faces (suggesting a role of holistic processing). Despite the prevalence of VO across many domains, a number of researchers have failed to replicate the phenomenon. However, a metaanalysis including 29 studies and 2,018 participants found a statistically significant overall effect, indicating that participants who verbalized a target were 1.27 times more likely than nonverbalizers to misidentify the target. Furthermore, this meta-analysis also revealed that verbalization effects are more likely to be observed when (a) the verbalization instructions induce detailed and elaborate descriptions and (b) there is a minimal time delay between the verbalization and the recognition test.

Currently, it is not entirely clear why VO occurs, and there are three competing accounts proposed to explain the phenomenon. The content account asserts that it is the precise contents of what is said during verbalization that interferes with memory. In other words, people create verbalizations that do not quite match up with the original visual memory, thus interfering with future recognition. The content account fits well with the effect of expertise on VO (see above) and helps explain why extensive verbalizations (which tend to include more inaccuracies) are more disruptive, but it fails to account for why many studies have failed to show a connection between the quality of verbal descriptions and recognition accuracy. It also has trouble accounting for why describing one face can interfere with the recognition of a different face. According to the criterion shift account, verbalization biases participants toward the target-“not-present” option, leading to reduced recognition accuracy when the target is present. Although this account potentially explains VO results in studies with “not-present” options, it fails to account for the many times in which VO has been observed in the absence of a not-present option. Finally, according to the process shift account, verbalization changes the individual’s processing orientation from a more holistic strategy to a more local one. If, as is often the case, faces are initially encoded holistically, the verbally induced featural processing strategy may lead to recognition processes that are incommensurate with the manner in which the face was originally encoded. A particular strength of the processing account is that it does not require a link between verbalization quality and recognition accuracy, which as stated above, is often not demonstrated. Although at present there is some controversy regarding the relative merits of the content, criterion, and processing shift accounts of VO, it seem likely that all three accounts have merit. Further research is needed to determine when each one is most likely to apply.

In summary, a large amount of research has shown VO to be a pervasive effect across many types of visual memory and other areas of perception. A recent meta-analysis has shown VO to be a relatively small, but reliable effect on memory. Furthermore, characteristics of the verbalizer, the particulars of the description task, and what occurs between viewing the original stimulus and retrieval of this stimulus appear to play a role in the strength of the VO effect. Although the exact cause(s) of VO remain unclear, that the effect occurs and that it can significantly effect eyewitness memory is beyond dispute. From a practical perspective, the modest size of VO effects combined with its tendency to dissipate over time suggests that investigators should not avoid soliciting verbal descriptions from witnesses. However, investigators may want to avoid asking participants to make identifications immediately after providing extensive verbal descriptions of a perpetrator’s face or voice.

References:

  1. Chin, J., & Schooler, J. W. (in press). Why do words hurt? Content, process, and criterion shift accounts of VO. European Journal of Cognitive Psychology. Available at:http://www.tandfonline.com/doi/abs/10.1080/09541440701728623
  2. Meissner, C. A., & Brigham, J. C. (2001). A meta-analysis of the VO effect in face identification. Applied Cognitive Psychology, 15, 603-616.

See also: