عدم تقارن نیمکره در پردازش معنایی: شواهدی از حافظه کاذب برای کلمه مبهم

Previous research suggests that the left hemisphere (LH) focuses on strongly related word meanings; the right hemisphere (RH) may contribute uniquely to the processing of lexical ambiguity by activating and maintaining a wide range of meanings, including subordinate meanings. The present study used the word-lists false memory paradigm [Roediger, H. L. III., & McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 803–814.] to examine whether these differences between the two cerebral hemispheres in semantic processing also affect memory representations for different meanings of ambiguous words. Specifically, we tested the differences between the LH and RH in recollecting unpresented, semantically related, ambiguous words following the presentation of lists of words all related to either the dominant or the subordinate meanings of these ambiguous words. Findings showed that for the unpresented ambiguous words, the LH made more false alarms than the RH for the dominant lists, whereas the opposite pattern emerged for subordinate lists. Moreover, d′ analyses showed that, whereas the LH was more sensitive to subordinate than dominant meanings, the RH showed no differences in sensitivity for the two types of word–lists. Taken as a whole, these results support the RH coarse semantic coding theory [Beeman, M. (1998). Coarse semantic coding and discourse comprehension. In Beeman & M., Chiarello, C. (Eds.), Right hemisphere language comprehension: Perspectives from cognitive neuroscience (pp. 255–284). Mahwah, NJ: Erlbaum; Jung-Beeman, M. (2005). Bilateral brain processes for comprehending natural language. Trends in Cognitive Sciences, 9, 512–518.] indicating that during word recognition, the RH activates and maintains a broader and less differentiated range of related meanings than the LH, including both dominant and subordinate meanings of ambiguous words. Furthermore, the findings suggest that hemispheric differences in ambiguity resolution during language processing extend also to verbal memory.

The accumulated evidence from neurologically intact, split-brain and brain-injured participants indicates that, although both cerebral hemispheres have access to word meanings, comprehension of semantic relations differs in the left (LH) and right (RH) hemispheres (for reviews, see e.g., Chiarello, 2003 and Jung-Beeman, 2005). Much research indicates that when a word is recognized by the LH, only the most strongly related meanings are activated, whereas in the RH a much broader set of meanings, including distant, unusual, non-salient, subordinate and figurative meanings becomes available (e.g., Beeman, 1998, Chiarello, 1991, Chiarello, 1998, Chiarello, 2003, Faust and Lavidor, 2003 and Jung-Beeman, 2005). One aspect of the qualitative differences between the hemispheres in semantic processing, which has been studied extensively in word recognition and comprehension, is the unique RH involvement in processing alternate meanings of ambiguous words (e.g., Burgess and Simpson, 1988, Coney and Evans, 2000, Faust and Chiarello, 1998 and Faust and Lavidor, 2003). The aim of the present study was to examine whether hemispheric differences in the ability to activate and maintain multiple meanings of ambiguous words, including subordinate, weakly related meanings holds also for verbal memory. Specifically, we used a modified word-lists false memory paradigm (e.g., Gallo, 2006, Howe, 2006 and Roediger and McDermott, 1995) to investigate the susceptibility of the LH and RH to unstudied lexically ambiguous words following the presentation of lists of words all related to either the dominant or subordinate meanings of these ambiguous words. 1.1. Hemispheric differences in semantic processing: The fine–coarse coding theory (FCT) An explanation of the role of the RH in lexical semantic processing has been provided by the fine–coarse semantic coding theory (FCT) developed by Beeman (e.g., Beeman, 1998 and Jung-Beeman, 2005). According to the FCT, immediately after encountering a word, the LH engages in relatively fine semantic coding, strongly and quickly focusing semantic activation on features related to the dominant, literally or contextually relevant meanings, while inhibiting features related to the subordinate or contextually irrelevant meanings. However, the RH engages in coarse semantic coding, weakly and diffusely activating large semantic fields containing multiple alternative meanings and more distant associates, including subordinate meanings of ambiguous words. Jung-Beeman (2005) has suggested that, while the strong categorical semantic activation in the LH is conducive to most language comprehension tasks, the biggest advantage conveyed by RH coarse semantic coding arises when people process multiple distantly related words. This is because given multiple input words, large semantic fields are more likely to overlap than smaller, more focused semantic fields. The findings of a multiple priming study (Beeman et al., 1994) supported this claim by showing that when people are presented with three-word primes (food, glass, pain) where each word is distantly related to the target word (cut), weak semantic activation from the three prime words summates in the RH, yielding stronger priming for the left visual field (LVF)/RH than for right visual field (RVF)/LH presented target words. However, a single, strongly related prime word (scissors) yielded stronger priming for RVF/LH than for LVF/RH presented target words. Thus, in response to multiple words, the RH’s large semantic fields are more likely to overlap than the LH’s small semantic fields. As a result, the RH is more likely to activate a concept that inferentially connects distantly related words ( Jung-Beeman, 2005). The notion that the LH and RH activate and maintain different word meanings has to be understood within the context of processing systems in which the availability of different types of information change over time (Chiarello, 2003). Previous priming research (e.g., Anaki et al., 1998 and Burgess and Simpson, 1988) generally indicates that the LH may initially activate a wide set of word meanings, but that this early stage is followed by a selection process in which the strongly related, dominant meanings are selectively maintained, while other weakly related meanings are discarded and not maintained for later processing. In contrast, the RH may be slower in meaning activation, but continues to maintain more distant meanings, including subordinate meanings, during time periods when these meanings are no longer available within the LH (for reviews see Beeman, 1998 and Chiarello, 2003). Hemispheric differences in the availability of alternate word meanings during different stages of language processing were examined in several priming studies by varying the interval (SOA) between the presentation of the prime and target word. Burgess and Simpson (1988) examined priming for both the dominant and subordinate meanings of ambiguous words (bank). At a short SOA of 35 ms, the less frequent subordinate meanings (river) were primed only within the RVF/LH, whereas at a 750-ms interval they remained accessible only within the LVF/RH. Thus, in the LH, only the more frequent dominant meanings were still active after relatively long intervals. In another study, Anaki et al. (1998) examined priming for laterally presented target words related to either the metaphorical weakly related (insult) or literal, strongly related (mosquito) meanings of centrally presented prime words that had both a literal and a metaphorical meaning (stinging). Bilateral priming of the metaphorical meanings was obtained at the short SOA (200 ms), but at a longer interval (800 ms) the metaphorical, more distantly related meanings were available only within the LVF/RH, whereas the literal, strongly related meanings were maintained only in the LH. The findings of both studies, as well as those of additional research (for review, see Chiarello, 2003), strongly suggest that the LH suspends rather rapidly processing of distantly related words, maintaining only the strongly related meanings for relatively long periods, while the RH functions to maintain a wide range of meanings for relatively long periods of time. 1.2. Hemispheric differences in true and false memories The differences between the LH and RH in the time course of meaning activation and maintenance suggest that the hemispheres may also fundamentally diverge in their memory representations of the alternate meanings of words. However, hemispheric differences in memory tasks remain understudied, particularly in non-clinical populations (Federmeier & Benjamin, 2005). A few studies have examined hemispheric differences in semantic memory using the word-lists false memory paradigm (Fabiani et al., 2000, Ito, 2001 and Westerberg and Marsolek, 2003). This paradigm, originally developed by Deese (1959) and later revived by Roediger and McDermott (1995), is currently referred to as the DRM (Deese/Roediger–McDermott) paradigm. In a typical experiment, participants are presented with 12-word–lists of thematically related words (e.g., blanket, bed, night) all associated with one non-presented critical lure (SLEEP). Subsequently, participants demonstrate a surprisingly high rate of false recall and false recognition for the critical lures. Roediger and McDermott (1995) successfully replicated Deese’s (1959) paradigm and found that the probability of recalling the critical lure approximated and even exceeded the probability of recalling words presented in the middle of the list. These findings have since been extensively replicated (e.g., Cabeza and Lennartson, 2005, Dibiberto-Macaluso, 2005, Gallo, 2006, Gallo et al., 2001, Hancock et al., 2003 and Smith et al., 2002) even when participants are forewarned about the purpose of the study ( Gallo et al., 1997 and McDermott and Roediger, 1998). The false-memory phenomenon demonstrated with the DRM paradigm, e.g., high rates of false recall and false recognition to the unpresented, critical words, is remarkable because the word–lists are short, memory tests follow learning immediately, and forward explicit warning of the phenomenon does not eliminate it (e.g., Gallo et al., 1997, McDermott and Roediger, 1998 and Watson et al., 2004). Although findings obtained with the DRM paradigm reveal similar memory levels for true and false memories, recent cognitive neuroscience studies have demonstrated marked differences in brain activity for true and false memories. Several studies have used functional neuroimaging techniques in attempts to identify specific brain regions associated with true and false memories. Using positron emission tomography (PET) scans, Schacter et al. (1996) found higher activation during true than false recognition in a temoropartietal region previously associated with auditory processing and retention. More recent studies have applied functional magnetic resonance imaging (fMRI) to test brain activity for true and false memories (Cabeza et al., 2001 and Okado and Stark, 2003). Using the DRM paradigm, Cabeza et al. (2001) documented a dissociation between two regions within the medial temporal lobe (MTL): the parahippocampal gyrus, which showed higher activity during true than false recognition, and the hippocampus, which showed significant activation during both true and false recognition. In line with these findings, other studies using fMRI have also found differential brain activity for true and false recognition for imagined words (Okado & Stark, 2003) and abstract shapes (Slotnick & Schacter, 2004). Finally, event-related potentials (ERPs) studies have revealed greater true than false recognition parietal effect (Curran et al., 2001 and Fabiani et al., 2000). Neuroimaging studies have also shown laterality effects in memory processes, including false memories. For instance, McDermott, Peterson, Watson, and Ojemann (2003) have found that encoding DRM materials activates regions in the left ventrolateral prefrontal cortex. This brain area was found to contribute to the cognitive control that permits strategic access to memory. Thus, in a recent review paper, Badre and Wagner (2007) show the contribution of these LH control mechanisms across a wide range of memory processes and tasks, supporting the notion that semantic access and retrieval processes may differ in the RH and LH. The most empirically supported explanation for the false-memory phenomenon in word–lists is the implicit associative response (IAR) theory (e.g., Hancock et al., 2003, Howe, 2006 and Underwood, 1965), according to which presentation of a list item not only activates the item itself, but also generates activation of semantic associates to that item, including the critical lure, resulting in later memory for the critical lure. More recently, McDermott and Watson, 2001 and Roediger et al., 2001, and Gallo and colleague (Gallo, 2006 and Gallo and Roediger, 2002) have presented an alternative explanation for how the information that leads to DRM false remembering is represented in memory, namely the activation/monitoring theory. According to this theory, processing the list items activates the critical non-presented lure, and false remembering reflects a failure to correctly monitor the source of this activation. Therefore, both the IAR and the activation/monitoring accounts hold that once an item is activated it can be stored into a longer lasting episodic memory, causing false recognition at longer delays. 1.3. DRM and laterality The DRM paradigm has been adopted also in the study of hemispheric differences in activation and maintenance of semantic information since it involves processes of semantic activation. As mentioned above, few previous studies (Fabiani et al., 2000, Ito, 2001 and Westerberg and Marsolek, 2003) have applied the DRM paradigm to demonstrate hemispheric differences in the processing of true and false memory, revealing inconsistent findings. Two studies have used a similar procedure (Ito, 2001 and Westerberg and Marsolek, 2003). In these studies, word–lists were presented in the study phase followed by a visual recognition test in which studied and critical words were presented to the LH or RH. However, whereas in Ito (2001) study, words in the study phase were presented visually, in Westerberg and Marsolek’s (2003) study, words in the study phase were presented auditorilly. These two studies revealed somewhat different findings: Based on hemispheric differences in the hit rates, Ito did not find any hemispheric differences in false memories (although the LH was more accurate than the RH in true memory retrieval). However, Westerberg and Marsolek actually found that the ability to correctly reject critical lures was better and participants were more confident in their rejection when test items were presented to the LH than to the RH. Fabiani et al. (2000) used ERP measures, in addition to a visual recognition test, to study lateralization effects for false memories with the DRM procedure. However, in Fabiani’s study, unlike Ito, 2001 and Westerberg and Marsolek, 2003 studies, words were presented laterally in the study phase and centrally in the test phase. Both the behavioral and electrophysiological measures indicated that the RH is less susceptible to semantically related unpresented lures. Fabiani et al.’s (2000) findings are in similar to those obtained by Metcalfe, Funell, and Gazzaniga (1995). Metcalfe et al. tested true and false memories for words from different categories in a patient who underwent a complete corpus-callosum resection. They found that the RH is better in rejecting unpresented lures. Some of the differences in the results of all these studies that examined lateralization effects for false memories can be attributed to differences in experimental design (e.g., mode of stimuli presentation). However, the general picture emerging from these studies strongly suggests that there are significant differences in how verbal material is retained by the two hemispheres that requires further research. Most of the studies that examined hemispheric differences in false memories for unpresented lures refer to the FCT (Beeman, 1998 and Jung-Beeman, 2005) in interpreting their findings. These studies have suggested that the model of RH coarse semantic coding and LH fine semantic coding may be applicable not only to language processing but also to verbal memory. Nevertheless, none of the studies manipulated the relative semantic distance of the presented words and critical lures and, therefore, cannot serve as a direct test of the FCT with regard to semantic memory in the RH and the LH. To directly test the validity of the FCT for hemispheric differences in semantic memory, in the present study, we manipulated the range of semantic activation using dominant vs. subordinate associates to ambiguous unpresented lures. Specifically, for each laterally presented unstudied ambiguous critical lure (e.g., second) we developed two word–lists, one associated with its dominant meaning (e.g., moment, time, watch) and the other associated with its subordinate meaning (e.g., first, last, line). In doing so, we could directly test the effect of dominant, closely related vs. subordinate, more weakly related word–lists on false memories for the unstudied ambiguous words in each hemisphere. The present study’s paradigm was similar to that of two previous studies that looked at the effect of multiple word priming on the recognition of ambiguous target words by the LH and RH (Faust and Kahana, 2002 and Faust and Lavidor, 2003). In both studies, participants saw two or three centrally presented priming words that were related to either the dominant or subordinate meaning of a laterally presented ambiguous target word. The experimental tasks included lexical decision and semantic judgments to the presented ambiguous target words. The findings of both studies indicated that the LH mostly benefited from multiple primes that converged onto the dominant meaning of the ambiguous target word, whereas the RH benefited most from several priming words that diverged on alternate meanings of the ambiguous target word. These results were interpreted as supporting the FCT by showing that during word recognition, the LH activates a restricted semantic field that includes mainly strongly related, dominant semantic features, whereas the RH activates a broader range of related meanings than the LH, including alternate meanings of ambiguous words. 1.4. Aim of present study The present study aimed to extend these findings to verbal memory using the DRM paradigm. The DRM paradigm is similar to the multiple priming paradigm in combining multiple words related to either the dominant or subordinate meaning of an ambiguous target (critical) word. However, it enables testing the maintenance of different meanings in each hemisphere over much longer periods of time. Based on the fine/coarse semantic coding in the LH and RH, respectively (Beeman, 1998 and Jung-Beeman, 2005), and on the results of previous priming studies (Faust and Kahana, 2002 and Faust and Lavidor, 2003), we hypothesized that, whereas the LH will show higher rates of false memories (i.e., lower sensitivity) for dominant- than subordinate-meaning lists, in the RH there would be no differences in false memories between the two types of lists. Furthermore, the RH may even falsely recognize weakly related lures more than strongly related ones (see, e.g., Anaki et al., 1998). Such findings would suggest that the differences between the RH and LH in semantic activation and maintenance generalize to memory.