«By Zachary Alexander Rosner A dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Psychology ...»
Generation also increased activation in regions involved in image generation (ITG) and object processing (LOC) (D’Esposito et al., 1997; Malach et al., 1995). Further, memory performance was correlated with increased generate activity in the PHG, temporal fusiform cortex, MTG, AG, LOC, and PrC, regions known to be important for memory binding and retrieval (Davachi, 2006), and the behavioral benefit (generate HC hit rate – read HC hit rate) of generating at 59 encoding was correlated with activity in medial anterior PFC regions known to be important for attending to internally generated versus externally perceived stimuli (Lagioia et al., 2011;
Simons et al., 2006, 2008). The DMN was also involved, as different regions within the DMN were active when reading or generating items during encoding (IPL, PrC, dMPFC for generate read, HC hits; IPL, PrC for read generate, HC hits). It is possible that on some trials, active generation oriented participants to internally generated information arising from semantic analysis or conceptual processes, while reading kept participants less on-task and allowed for increased mind wandering (see Buckner & Carroll, 2007; Shimamura, 2011; Spreng et al., 2009).
Interestingly, the successful retrieval effect (hits CRs at retrieval), which was associated with activation in lateral and medial PPC (see Cabeza, 2008; Shimamura, 2011; Vilberg & Rugg, 2008), revealed no significant differences for items that were initially read or generated, suggesting that the benefits of generation act at encoding. These findings link the generation effect to regional activations during encoding that are known to be critical for the establishment of long-term memories (Paller & Wagner, 2002). Generation increased both prefrontal activity and activity in posterior regions involved in verbal processing, object analysis, and visuospatial imagery.
Even with the praise of educators and great efforts of psychological researchers, the exact reason as to why active generation improves memory remains unknown. Various theories have suggested that active generation increases cognitive effort (McFarland Jr. et al., 1980; Tyler et al., 1979), conceptual processing (Jacoby, 1983), item distinctiveness (Begg et al., 1989; Hunt & McDaniel, 1993; Kinoshita, 1989), semantic processing (McElroy, 1987; McElroy & Slamecka, 1982), the association between cue and target items (Hirshman & Bjork, 1988), and the relationship between target items (McDaniel et al., 1988). And, while the positive effects of generation are robust, there are limitations. The generation effect is reduced when the read and generate conditions are manipulated between rather than within subjects (Slamecka & Katsaiti, 1987), and vanishes when using stimuli without preexisting representations (Gardiner & Hampton, 1985; Nairne et al., 1985).
Perhaps the reason that there has been no consensus as to which of these presented theories drives the generation effect is because the effects of active generation are ubiquitous and engage a large range of cognitive processes. With respect to the fMRI findings, it is clear that multiple brain regions are responsible for different aspects of the mnemonic benefit associated with the generation effect. It is possible that active generation increases attention and cognitive effort (prefrontal and posterior cortical activation; Miller & Cohen, 2001; Shimamura, 2000, 2008), conceptual and semantic processing (IFG and MTG; Bookheimer, 2002; Poldrack et al., 1999), and item distinctiveness (LOC and ACC; Malach et al., 1995; van Veen et al., 2001). The benefits of active generation during encoding likely depend on the nature of the retrieval task, supporting a TAP account (see Blaxton, 1989; Jacoby, 1983). Essentially, the amount of overlap between the processes preferentially engaged by generation at encoding with later retrieval will drive the mnemonic benefit. Various encoding tasks can bias the engagement of relevant cognitive processes, promoting relative increases in attention, cognitive effort, item distinctiveness, semantic processing, and conceptual processing.
In spite of these results, active learning does not simply enhance memory across the board. Generation has been demonstrated to impair memory for contextual information such as temporal order (Nairne et al., 1991), color (Mulligan, 2004; Mulligan et al., 2006), and the person who presented information (Jurica & Shimamura, 1999). As with the positive generation effect, the results of this negative generation effect have been inconsistent. Location memory, for 60 example, has had generation effects varying from positive to null (Marsh, 2006; Mulligan et al., 2006). Various theories have been proposed to explain these negative generation effects, or lack thereof, including item-context tradeoff accounts (Jurica & Shimamura, 1999; Nairne et al., 1991), TAP accounts (Mulligan, 2004, 2011; Mulligan et al., 2006), and accounts in which generation can enhance item-context relationships (Marsh et al., 2001).
The present experiments demonstrate negative generation effects not only for context memory, but also for item memory under certain conditions. When participants were presented with picture stimuli at both encoding and retrieval, generation impaired memory for item information. In addition, the Category Retrieval Blocking experiment demonstrated that while active generation enhanced memory for the generated items themselves, this benefit impaired memory for related non-generated items. This effect may operate in a way similar to that of retrieval-induced forgetting (Anderson et al., 1994; Bäuml, 2002). The Idiom experiments further demonstrated that generation negatively impacts color memory while failing to influence background color memory and location memory (see Mulligan, 2004; Mulligan et al., 2006).
Interestingly, however, the cross-cultural Idiom experiments revealed cultural differences for the way in which context memory may be processed. While active generation impaired color memory and left location memory unaffected among American participants, among Chinese participants it impaired both.
Taken together, this collection of experiments supports a TAP account as adopted by Mulligan (2011). In this view, generation enhances memory for item information through increased conceptual processing, impairs memory for intrinsic contextual details through increased perceptual processing, and has no effect on memory for extrinsic contextual details. As seen in the fMRI experiment, active generation can recruit a broad range of cognitive processes, and its benefit depends on the overlap between these processes and those engaged during later retrieval. Generation effect experiments are typically constructed in such a way that generation promotes conceptual processing, while passive study promotes perceptual processing. As most standard tests of item recall and recognition are considered to be conceptual in nature, generation often benefits item memory. Consistent with this theory, the American Idiom experiments demonstrated a positive generation effect for item memory, a negative generation effect for text color memory (intrinsic contextual feature), and no generation effect for location memory (extrinsic contextual feature) or background color memory (extrinsic contextual feature). In contrast, Chinese participants maintained a positive generation effect for item memory and a negative generation effect for text color memory (still an intrinsic contextual feature) while demonstrating a negative generation effect for location memory (now an intrinsic contextual feature). As Chinese participants are more likely than American participants to process location as an intrinsic contextual detail due to a field-dependent cognitive style (Markus & Kitayama, 1991), these results fit comfortably within the proposed framework.
Further, the Picture Fragment Completion experiments demonstrated an instance of a negative generation effect on item memory. In these experiments, participants were tested with pictures, a recognition task that is more perceptual in nature than typical word recognition tasks.
However, the type of stimuli presented during encoding impacted the way in which item information was accessed at test. When presented with pictures at study, participants capitalized on the consistent perceptual processing afforded by passive study, resulting in a negative generation effect for both item and color memory. When presented with words at study, however, this advantage of consistent perceptual processing of item information between study and test vanished, requiring participants to rely on conceptual features of the items instead.
61 Therefore, a positive generation effect on item memory was found, but was accompanied by a negative generation effect for color memory, as conceptual processing was likely of little utility for this task.
When considering this entire collection of results, Mulligan’s (2011) TAP account appears to be the most viable explanation for the various positive and negative effects of generation. Negative generation effects were found for item, color and location memory, disputing an account that generation enhances item and context memory (Marsh et al., 2001).
Under different circumstances, however, it is certainly plausible that active generation could benefit context memory to the extent that this information is targeted by the act of generation itself. Additionally, positive generation effects for item memory were paired with both negative and null generation effects for context memory in the Idiom experiments. The Picture Fragment Completion studies, on the other hand, paired a negative generation effect for item memory with a negative generation effect for context memory in the Picture-Picture experiment and a positive generation effect for item memory with a negative generation effect for context memory in the Word-Picture experiment. Given that positive and negative item and context memory generation effects could be independently manipulated, it appears unlikely that the negative generation effect for context memory is a necessary consequence of the positive generation effect for item memory. While these results argue against a strict tradeoff account (Jurica & Shimamura, 1999), it is likely that the majority of generation effect experiments do result in the processing of one type of information (i.e., conceptual processing or item information) at the expense of processing other types of information (i.e., perceptual processing or contextual information).
While the studies presented in this dissertation, backed by nearly 40 years of research showing the generation effect to yield nearly a 10% memory benefit (Bertsch et al., 2007), demonstrated the power of the generation effect through the use of synonyms, antonyms, idioms, categories and pictures, the utility of active generation as an encoding strategy outside of the laboratory is questionable. Generating the second word of a synonym pair or the last word of an idiom offers little value to the student who wants to learn a foreign language, understand history, or study for a science test. Therefore, future experiments should investigate ways in which active generation may be employed to enhance learning in real-world settings such as textbooks and classrooms.
While students may not dispute the value of active learning, Karpicke et al., (2009) found that college students tend to reread textbooks or notes rather than employ active learning strategies such as self-testing. To be sure, self-testing has great value. Self-testing allows students to recognize when they have sufficiently learned information (Karpicke and Roediger,
2008) and its memory benefit outweighs the benefit of repeated study sessions (Agarwal et al., 2008; Roediger III & Karpicke, 2006a), elaborative encoding (Karpicke & Blunt, 2011), and even active generation (Karpicke & Zaromb, 2010). While self-testing is powerful, the investigation of the generation effect in the classroom warrants attention. Generation activities can be incorporated into study materials more easily than self-testing, prompting more widespread use in studying. Active generation is also a useful strategy for initially encoding information, while self-testing can only capitalize on previously learned information.
Indeed, studies have found that generation enhances learning in education, and the errors students might make by generating incorrect information are not harmful if feedback is provided to correct those errors (Metcalfe & Kornell, 2007). Further, the benefits of generation can continue past the initial encoding phase, as students who generated words a second time showed a memory advantage over those who generated and then read words or simply read words twice 62 (MacLeod et al., 2012). Additionally, as students seem hesitant to take the initiative to test themselves, active generation can be easily adopted into classroom activities, textbooks, and study materials, especially during a time in which electronic resources are becoming more readily available. For example, answers to facts could be incomplete, bilingual word pairs could require generation, and key terms in textbooks could be fragmented. The areas of study in which active generation can enhance learning should be exhausted. Additionally, the optimal conditions of active learning should be investigated. Should people read first, followed by generation, followed by self-testing? Do these strategies depend on the type of the to-be-learned material?
Throughout this collection of experiments, active generation proved to be a powerful encoding strategy, engaging a wide range of cognitive processes and broad networks of brain activity. Seemingly an almost effortless task, generation enhanced item memory for various stimuli under several conditions. Active generation, however, can have limitations and consequences, as it was demonstrated to impair both item and context memory in certain situations. A TAP account (Mulligan, 2011) reasonably accounts for this pattern of positive and negative generation effects.
63 Chapter 6: References Agarwal, P. K., Karpicke, J. D., Kang, S. H. K., Roediger III, H. L., & McDermott, K. B. (2008).
Examining the testing effect with open- and closed-book tests. Applied Cognitive Psychology, 22(7), 861–876.
Anderson, M. C., Bjork, E. L., & Bjork, R. A. (2000). Retrieval-induced forgetting: Evidence for a recall-specific mechanism. Psychonomic Bulletin & Review, 7(3), 522–530.
Anderson, M. C., Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause forgetting:
Retrieval dynamics in long-term memory. Journal of Experimental Psychology: