Bastin et al. make a valiant attempt to rejuvenate the dual-process theory of recognition – a theory formulated more than 40 years ago, and which has triggered intense debate and research ever since (Dunn Reference Dunn2004; Wixted Reference Wixted2007; Wixted & Mickes Reference Wixted and Mickes2010; Yonelinas Reference Yonelinas2002). Their attempt yields the integrative memory model. The most novel and significant aspect of this model is the clear distinction between mnemonic representations – entity and relational – and the subjective experiences – familiarity and recollection – that are attributed to reactivation of the corresponding representations. This distinction has been misleadingly blurred in previous dual-process model of recognition. According to the integrative memory model, familiarity is typically associated with entity representations and recollection is typically associated with relational representations. However, recollection and familiarity are not necessarily triggered by relational and entity representations, respectively. Thus, the model makes the novel prediction that the subjective experience of familiarity may also be triggered by relational representations and, likewise, the subjective experience of recollection may also be triggered by entity representations (though this latter point is only implied, and not explicitly mentioned in the target article). This is an important prediction – perhaps the most important prediction that the model makes. However, its implications are not fully explored by the authors and are not sufficiently elaborated on. In this commentary, I focus on one specific aspect of this prediction – namely, that reactivation of a trace in the relational core system may trigger familiarity.
Bastin et al. claim that “even if the relational representation core system reactivates specific item-context details, one may experience a feeling of familiarity” (sect. 3, para. 2). This claim begs the question: What gives rise to this feeling of familiarity? However, the model remains mute regarding the source for the experience of familiarity for relational representations. For entity representations, the source triggering familiarity is the fluency heuristic, defined as “the speed and ease with which a stimulus is processed” (sect. 2, para. 2). Importantly, while mentioning other factors that are potential sources of familiarity (e.g., proprioceptive and affective information), the authors acknowledge that these may also be intricately linked to fluency – either being a by-product of fluency, or by triggering fluency (see sect. 4.2.1). Thus, familiarity emerges predominantly from the fluency with which a stimulus is processed. In line with the vast majority of the relevant literature (e.g., Kleider & Goldinger Reference Kleider and Goldinger2004; Whittlesea et al. Reference Whittlesea, Jacoby and Girard1990; Whittlesea & Leboe Reference Whittlesea and Leboe2000), Bastin et al. describe fluency only as a property of processing entity representations (e.g., single words, pictures). However, I maintain that there is no reason to preclude fluency as a relevant property of relational representations as well. Hence, a feasible source for the subjective experience of familiarity for relational representations is the fluency with which these representations are processed. Two lines of evidence support this notion.
First, though fluency is typically examined in the context of single entities, effects of fluency have also been demonstrated for information which can be construed as relational. Such relational information may include a variety of content types which, critically, involve the binding of two or more items – namely, forming a link between the items while preserving the meaning of each individual item (Eichenbaum et al. Reference Eichenbaum, Otto and Cohen1994). For instance, processing of arithmetic problems, which are comprised of several numbers and the relations between them, is affected by fluency (manipulated as the number of exposures to each problem; Paynter et al. Reference Paynter, Reder and Kieffaber2009; Reder & Ritter Reference Reder and Ritter1992). Sentences are an additional instance of relational information whose representations include both their constituent elements (namely, the words) and the relationships between them. Fluency for sentences has been manipulated both visually, by comparing sentences written in a degraded font to those written in a clear font (Alter et al. Reference Alter, Oppenheimer, Epley and Eyre2007; Laham et al. Reference Laham, Alter and Goodwin2009; Song & Schwarz Reference Song and Schwarz2008a), and aurally, by comparing sentences pronounced with a non-native accent versus a native accent (Lev-Ari & Keysar Reference Lev-Ari and Keysar2010). Interestingly, the effects of the fluency with which sentences got processed were demonstrated on various dependent measures, including syllogistic reasoning, speakers’ credibility, and moral judgments. Going beyond sentences, fluency has been shown to affect processing of whole paragraphs (Diemand-Yauman et al. Reference Diemand-Yauman, Oppenheimer and Vaughan2011; Song & Schwarz Reference Song and Schwarz2008b). Finally, fluency has also been shown to exert its effects on processing of ambiguous paintings which, as in previous examples of relational information, are composed of several elements and the relationships between them (Jakesch et al. Reference Jakesch, Leder and Forster2013).
A second line of evidence supporting the relevance of fluency to relational representations concerns repetition suppression – reduction of neural activity to repeated presentations of stimuli. Repetition suppression is regarded by many (apparently, Bastin et al. included) as a neural marker for fluency (e.g., Ward et al. Reference Ward, Chun and Kuhl2013). The target article describes several pieces of evidence for repetition suppression in the perirhinal cortex, a key structure in the entity representation core system. However, repetition suppression is not limited to neural structures within the entity representation system, and has also been demonstrated for relational representations in the hippocampus (Duncan et al. Reference Duncan, Ketz, Inati and Davachi2012; Düzel et al. Reference Düzel, Habib, Rotte, Guderian, Tulving and Heinze2003; Kumaran & Maguire Reference Kumaran and Maguire2006; Reference Kumaran and Maguire2007; Reference Kumaran and Maguire2009). For instance, in one study relational representations were operationalized as face–object and face–location associations (Düzel et al. Reference Düzel, Habib, Rotte, Guderian, Tulving and Heinze2003). Decreased activity in the hippocampus was found for repeated associations (intact pairs) versus novel associations (recombined pairs). Thus, the hippocampus – a key structure in the relational representation core system – also exhibits repetition suppression, the neural correlate of fluency.
To conclude, the integrative memory model makes the novel prediction that familiarity can be attributed to relational representations. However, the model does not specify what would lead to this attribution. Therefore, an exciting avenue for future research is to elucidate the sources of information or heuristics that may give rise to familiarity for relational representations. Based on the two lines of evidence reviewed above, I suggest considering fluency as a major candidate.
You have
Access
Bastin et al. make a valiant attempt to rejuvenate the dual-process theory of recognition – a theory formulated more than 40 years ago, and which has triggered intense debate and research ever since (Dunn Reference Dunn2004; Wixted Reference Wixted2007; Wixted & Mickes Reference Wixted and Mickes2010; Yonelinas Reference Yonelinas2002). Their attempt yields the integrative memory model. The most novel and significant aspect of this model is the clear distinction between mnemonic representations – entity and relational – and the subjective experiences – familiarity and recollection – that are attributed to reactivation of the corresponding representations. This distinction has been misleadingly blurred in previous dual-process model of recognition. According to the integrative memory model, familiarity is typically associated with entity representations and recollection is typically associated with relational representations. However, recollection and familiarity are not necessarily triggered by relational and entity representations, respectively. Thus, the model makes the novel prediction that the subjective experience of familiarity may also be triggered by relational representations and, likewise, the subjective experience of recollection may also be triggered by entity representations (though this latter point is only implied, and not explicitly mentioned in the target article). This is an important prediction – perhaps the most important prediction that the model makes. However, its implications are not fully explored by the authors and are not sufficiently elaborated on. In this commentary, I focus on one specific aspect of this prediction – namely, that reactivation of a trace in the relational core system may trigger familiarity.
Bastin et al. claim that “even if the relational representation core system reactivates specific item-context details, one may experience a feeling of familiarity” (sect. 3, para. 2). This claim begs the question: What gives rise to this feeling of familiarity? However, the model remains mute regarding the source for the experience of familiarity for relational representations. For entity representations, the source triggering familiarity is the fluency heuristic, defined as “the speed and ease with which a stimulus is processed” (sect. 2, para. 2). Importantly, while mentioning other factors that are potential sources of familiarity (e.g., proprioceptive and affective information), the authors acknowledge that these may also be intricately linked to fluency – either being a by-product of fluency, or by triggering fluency (see sect. 4.2.1). Thus, familiarity emerges predominantly from the fluency with which a stimulus is processed. In line with the vast majority of the relevant literature (e.g., Kleider & Goldinger Reference Kleider and Goldinger2004; Whittlesea et al. Reference Whittlesea, Jacoby and Girard1990; Whittlesea & Leboe Reference Whittlesea and Leboe2000), Bastin et al. describe fluency only as a property of processing entity representations (e.g., single words, pictures). However, I maintain that there is no reason to preclude fluency as a relevant property of relational representations as well. Hence, a feasible source for the subjective experience of familiarity for relational representations is the fluency with which these representations are processed. Two lines of evidence support this notion.
First, though fluency is typically examined in the context of single entities, effects of fluency have also been demonstrated for information which can be construed as relational. Such relational information may include a variety of content types which, critically, involve the binding of two or more items – namely, forming a link between the items while preserving the meaning of each individual item (Eichenbaum et al. Reference Eichenbaum, Otto and Cohen1994). For instance, processing of arithmetic problems, which are comprised of several numbers and the relations between them, is affected by fluency (manipulated as the number of exposures to each problem; Paynter et al. Reference Paynter, Reder and Kieffaber2009; Reder & Ritter Reference Reder and Ritter1992). Sentences are an additional instance of relational information whose representations include both their constituent elements (namely, the words) and the relationships between them. Fluency for sentences has been manipulated both visually, by comparing sentences written in a degraded font to those written in a clear font (Alter et al. Reference Alter, Oppenheimer, Epley and Eyre2007; Laham et al. Reference Laham, Alter and Goodwin2009; Song & Schwarz Reference Song and Schwarz2008a), and aurally, by comparing sentences pronounced with a non-native accent versus a native accent (Lev-Ari & Keysar Reference Lev-Ari and Keysar2010). Interestingly, the effects of the fluency with which sentences got processed were demonstrated on various dependent measures, including syllogistic reasoning, speakers’ credibility, and moral judgments. Going beyond sentences, fluency has been shown to affect processing of whole paragraphs (Diemand-Yauman et al. Reference Diemand-Yauman, Oppenheimer and Vaughan2011; Song & Schwarz Reference Song and Schwarz2008b). Finally, fluency has also been shown to exert its effects on processing of ambiguous paintings which, as in previous examples of relational information, are composed of several elements and the relationships between them (Jakesch et al. Reference Jakesch, Leder and Forster2013).
A second line of evidence supporting the relevance of fluency to relational representations concerns repetition suppression – reduction of neural activity to repeated presentations of stimuli. Repetition suppression is regarded by many (apparently, Bastin et al. included) as a neural marker for fluency (e.g., Ward et al. Reference Ward, Chun and Kuhl2013). The target article describes several pieces of evidence for repetition suppression in the perirhinal cortex, a key structure in the entity representation core system. However, repetition suppression is not limited to neural structures within the entity representation system, and has also been demonstrated for relational representations in the hippocampus (Duncan et al. Reference Duncan, Ketz, Inati and Davachi2012; Düzel et al. Reference Düzel, Habib, Rotte, Guderian, Tulving and Heinze2003; Kumaran & Maguire Reference Kumaran and Maguire2006; Reference Kumaran and Maguire2007; Reference Kumaran and Maguire2009). For instance, in one study relational representations were operationalized as face–object and face–location associations (Düzel et al. Reference Düzel, Habib, Rotte, Guderian, Tulving and Heinze2003). Decreased activity in the hippocampus was found for repeated associations (intact pairs) versus novel associations (recombined pairs). Thus, the hippocampus – a key structure in the relational representation core system – also exhibits repetition suppression, the neural correlate of fluency.
To conclude, the integrative memory model makes the novel prediction that familiarity can be attributed to relational representations. However, the model does not specify what would lead to this attribution. Therefore, an exciting avenue for future research is to elucidate the sources of information or heuristics that may give rise to familiarity for relational representations. Based on the two lines of evidence reviewed above, I suggest considering fluency as a major candidate.