If verbal rehearsal specifically interferes with the application of non-verbal knowledge, then its effects should generalize across different types of perceptual memories. Consistent with this prediction we have now found that verbal overshadowing generalizes to other visual memories such as visual forms (Brandimonte, Schooler, & Gabbino, 1997) and maps (Fiore and Schooler, in preparation) as well as other perceptual modalities including, audition (Schooler, Fiore, and Brandimonte, 1997; Fiore, Houser, and Schooler, 2001) and taste (Melcher & Schooler, 1996). In contrast, verbalizable memories are generally unaffected by verbalization, including memory for statements (Schooler & Engstler-Schooler, 1990) and easy to name forms (Brandimonte et al., 1997).

If verbalization interferes with the nonverbal components of memory, then the effects of verbalization should depend on one's relative degree of nonverbal versus verbal expertise. Accordingly, if domain-specific nonverbal expertise markedly exceeds verbal expertise, then performance should be impaired following verbalization. If, however, the two forms of expertise are more commensurate, then shifting between the two should be less problematic. We have documented this relationship between expertise and verbal overshadowing in both face recognition (Fallshore & Schooler, 1995; Ryan & Schooler, 1999) and wine recognition (Melcher & Schooler, 1996).

One recurring issue in interpreting memory interference effects is the fate of the disrupted memory (e.g., Loftus & Loftus, 1980). Is it lost or just temporarily inaccessible? We have now conducted several studies demonstrating that the effects of verbalization can be reversed by introducing retrieval conditions that encourage access to the original perceptual memory (e.g., Brandimonte, et al, 1997; Schooler & Engstler-Schooler, 1990; Schooler, Ryan, & Reder, 1996). Thus, verbalization overshadows but does not eradicate the original memory.

The most straightforward account of the negative effects of verbalization is that it involves a source monitoring error (Johnson, Hastroudi, & Lindsay, 1993) in which the non-veridical verbalizations are confused with the original memory. However, persistent efforts have failed to support such an account. For example, we consistently find no relationship between the contents of subjects’ verbalizations and the accuracy of their performance (Fallshore & Schooler, 1995; Schooler & Engstler-Schooler, 1990). Warning subjects to ignore the contents of their verbalizations, if anything, exacerbates verbal overshadowing effects (Dodson, Johnson, & Schooler, 1997). Moreover, verbalization of one target can actually interfere with the recognition of a different target (Dodson et. al, 1997). Thus, verbal overshadowing does not appear to result from a reliance on the actual descriptions but rather involves a more general disruption of non-reportable knowledge/processes (for further theoretical analysis see Schooler, Fiore, & Brandimonte, 1997).

Attitudes are notoriously difficult to explain (Nisbett & Wilson, 1977), thereby making attitudes potentially vulnerable to verbal analysis. Consistent with this prediction, Tim Wilson and I have found various situations in which verbally analyzing the basis for one’s attitudes can be detrimental. For example, Wilson and Schooler (1991) found that verbally analyzing one’s reasons for affective judgments (taste quality of strawberry jams) caused subjects to make decisions that were less in line with the opinions of experts, relative to nonverbalizing controls. Similarly, we found that verbally analyzing reasons causes individuals to make decisions that result in less post-choice satisfaction (Wilson, Lisle, Schooler, Hodges, Klaaren and Lafleur, 1993)

Insight problem solving represents another domain of cognition that is often alleged to involve non-reportable processes (e.g., Metcalfe & Weibe, 1987). Consistent with this view (and with the claim that verbalization disrupts non-reportable processes), we have found that verbalization (thinking out loud) markedly disrupts insight but not logical problem solving (Melcher & Schooler, 1995; Schooler, Ohlsson, & Brooks, 1990). This finding piqued my interest in the nature of non-reportable insight processes. Using a split visual field priming paradigm we found evidence that insight processes may be specifically associated with the right-hemisphere (Fiore & Schooler, 1997). Using an individual differences paradigm, we (Schooler & Melcher, 1995) found that insight problem solving may draw on processes similar to those involved in recognizing out-of-focus pictures (a task that we have also found to be vulnerable to verbalization). Collectively, these finding suggest that the non-reportable processes associated with insight may be similar to basic perceptual pattern recognition processes (cf. Schooler, Fallshore, & Fiore, 1995).

The appreciation of the deep structure mapping between superficially different problems is very challenging for many students. Indeed, even when individuals do recognize an analogical mapping, they often have difficulty articulating the basis of the analogy, (Gick and Holyoak, 1983). Thus, analogical judgments represent yet another domain in which verbalization might be expected to be disruptive. Consistent with this view, Sieck, Quinn, and Schooler (1999), found that requiring participants to verbally justify judgments of the soundness of analogies between stories reduced their ability to distinguish between pairs that involved true analogies (shared deep structure) and those involving only mere appearance analogies (shared surface structure). More recent findings indicate that the retrieval of true analogies is similarly disrupted by verbalization (Lane & Schooler, 2004)