Most of the previous research on the DRM paradigm has focused on lists of words. The current study extends the DRM paradigm to full sentences, and furthermore, complex sentences involving figurative language. An analogous memory illusion was found by Bransford and Franks (1971) in which non-presented sentences were falsely recognized if they contained highly similar semantic content to the presented sentence. In that study, however, the sentence lures were much more obviously related and shared many of the same words. In the current study, the sentence lures were less related, not sharing as much overlap in terms of words, and the sentences were not all referring to the same event or occurrence. However, the sentences were related in terms of a deeper underlying meaning, in this case sharing a common cross-domain metaphorical mapping. Therefore, this finding extends previous research on false recognition for sentences by suggesting that even when sentences are fairly unrelated on the surface, they can still induce false recognition if they are based on the same metaphorical mapping.
Although the goal of this thesis was not to test alternative false memory theories, I speculate on how three major theories, automatic spreading activation, processing
fluency, and fuzzy-trace theory, could accommodate the findings. Following this, I consider the findings from the “remember,” “know,” and “guess” judgments and how they may relate to CMT and DMT.
6.1.1
Automatic Spreading Activation
Roediger, Balota, and Watson (2001) suggest that the DRM effect is due to automatic associative activation and mainly focus on word-based similarity or associations, but in the current study, word-based similarity cannot fully account for the results. In fact, the critical and control lures did not differ significantly in terms of word similarity to their respective study lists, at least as measured in terms of a vector-based word model, which should tap into both similarity and association (Clark, 2015; Reid & Katz, 2018b). Therefore, it seems that false memory in the DRM paradigm can be elicited by factors beyond word similarity. In this case, it was elicited by metaphorical similarity.
Automatic spreading activation could accommodate the current findings if metaphorical expressions are stored in semantic memory somewhat like words (e.g., Swinney & Cutler, 1979; see Nayak & Gibbs, 1990, for a review of these accounts). Similar to how nodes for related words are connected, nodes for expressions might also be connected when they are based on the same underlying conceptual metaphor. That is, expressions that use the same cross-domain mapping are likely stored in the same semantic memory structure, or at least these expressions are in closer proximity than expressions using different cross-domain mappings. By this account, activation could possibly spread between nodes for expressions as it does with nodes for words. I repeat this is speculation because expressions really have not been considered in the spreading activation literature, especially regarding DRM false memory effects.
6.1.2
Processing Fluency
Processing fluency and spreading activation explanations of false memory go hand-in- hand as spreading activation results in fluent processing of related lures (Gallo & Roediger, 2003; Whittlesea, 2002). However, the critical difference is that processing fluency does not depend on metaphorical expressions being stored in semantic memory. Alternatively, conceptual metaphors may act more like schemas that help to organize and
interpret metaphorical expressions and the schema may be activated when a metaphorical expression is read. In other words, the expression itself is not retrieved from semantic memory, but the conceptual metaphor schema is employed to make sense of the
expression. This would still lead to the same processing fluency advantage because if the schema is already activated from reading prior expressions, new expressions should be more readily processed.
A processing fluency explanation could be tested using the paradigm employed by Whittlesea (2002), who had participants read DRM-type lists, but then make a lexical decision on the critical lure rather than a recognition judgment. Applied to the current study, participants would first read the study list expressions and then read related and unrelated critical consistent lure sentences. A processing fluency explanation would depend on the critical consistent lures being read faster when related to the study list than when unrelated. Some previous research has found that reading metaphor expressions based on the same conceptual metaphor facilitates processing of related expressions, but the findings are mixed, and either way, facilitated processing has not been confirmed with the stimuli employed in the current study.
6.1.3
Fuzzy-trace theory
Lastly, the conceptual metaphor false memory effect could be explained in terms of fuzzy-trace theory (Brainerd & Reyna, 2005). Recall that this theory focuses on the “gist,” or the deeper theme or meaning of the to-be-remembered information. From this perspective, the “gist” of each study list would essentially be the underlying conceptual metaphor, and the critical consistent lures would be consistent with the “gist” of each list. Although some have argued that gist extraction likely relies on more effortful processing, and therefore, more conscious awareness than a spreading activation explanation
(Carneiro, Garcia-Marques, Fernandez, & Albuquerque, 2014), Brainerd and colleagues (Brainerd & Reyna, 2005; Brainerd, Forrest, Karibian, & Reyna, 2006) do not make any claims about whether gist extraction is conscious or unconscious. Therefore, a gist-based explanation for the findings from this set of studies accommodates the data nicely.
The purpose of this thesis was not to test alternative theories of false memory, but to test assumptions of CMT, so I do not make strong claims regarding which memory theory best supports the data. Overall, the three outlined theories all reasonably accommodate the findings.
6.1.4
Remember, Know, and Guess Judgments
Across Chapters 3, 4, and 5 in which the remember-know-guess procedure was
employed, the results suggest that false recognitions of the critical consistent lures were attributed mostly to familiarity. Over all these chapters, the falsely recognized critical consistent lures never had a percentage of remember judgments above 50%, and the percentage of guess judgments was fairly high across all experiments, suggesting a weaker and less confident sense of familiarity (Hirshman, 1998). Although these data on their own cannot be used to confidently distinguish between CMT and DMT, when considered in light of the other findings from the strategy descriptions and divided attention manipulations, they further suggest that conceptual metaphors are activated automatically. Familiarity is considered to be the more unconscious and automatic type of memory whereas recollection is considered more effortful and intentional. Given the automaticity assumption of CMT, one would expect a greater degree of familiarity-based false recognitions. This is also consistent with the notion that conceptual metaphor activation facilitates processing of related expressions, as processing fluency primarily affects familiarity. In contrast, one may expect more false recollection if cross-domain mappings depend on conscious and deliberate processing, as proposed by DMT. I repeat that this cannot carry the argument for CMT over DMT; however, given the entirety of the data, the high percentage of false familiarity further supports the automaticity of conceptual metaphor activation.
It should be noted that the low percentage of “remember” responses could also be due to methodological factors unrelated to CMT and DMT. For instance, full sentences were employed, and participants may have hesitated to use “remember” judgments if any single word in the lure sentence seemed unfamiliar. Full sentences also require a longer presentation duration (3 seconds) which is a factor known to weaken false recognition effects (McDermott & Watson, 2001) and potentially resulted in less false “remember”
judgments in the current study. At best, the “remember,” “know,” and “guess” judgments should be interpreted cautiously, but the data is consistent with the original CMT.