선택한 단어 수는 입니다.
최소 단어 이상 선택하여야 합니다.
최소 단어 이상 선택하여야 합니다.
선택한 단어 수는 30입니다.
최대 10 단어까지만 선택 가능합니다.
최대 10 단어까지만 선택 가능합니다.
문제 정의
- In terms of subordinate information, it is not implausible that the categorization of episodic memories will cause disruption of the details of episodic memory, as less informative detailed memories would be crowded out when cognitive resources are demanded for more representative information; especially when one needs to create a category rather than choose from a limited number of existing ones. The aim of the present study was to investigate such disruptions by testing the effect of semantic categorization on episodic details of memory. We instated a new paradigm for our research that requires the encoding and retention of episodic details while ongoing categorization of superordinates.
가설 설정
- Our study is not without limitations. First, the time period of assumed categorization is relatively broad. A six-second-long Encoding detail task was implemented between two Categorization tasks, and we presumably define that gap as the time period that includes the moment of categorization.
제안 방법
- Each experimental trial was treated as an event of zero duration in the general linear model (GLM). All four runs were concatenated for conventional GLM analysis while considering the session effect from each run, and GLMs for the Categorization tasks and Encoding details tasks of the experimental trials were modeled separately in both experimental and control runs. Regressors 1 to 5 were the Categorization tasks of the first to fifth trials of the experimental runs, and 6 to 10 were the Encoding detail words tasks of the trials in the same order.
- , 2002; Raposo, Moss, Stamatakis, & Tyler, 2006; Wagner, Bunge & Badre, 2004) based on prior studies were the main focus of our study. Considering our findings in behavioral results, we contrasted the neural activities of tasks included in the trials in sequential order as we did in the behavioral analyses (i.e., tasks in first trials vs. tasks in second trials, so on), while focusing on the time period between the second and third trials as the possible moment of successful categorization. Changes of neural activations in target regions in each trial were consistent with the behavioral results.
- 78), t(17) = 0, p = 1. Considering the differences found in both RTs and confidence ratings, we postulate that categorization is successfully established while progressing from second to third trials. In the Encoding details tasks, RT only decreased significantly between the first (M = 2305.
- Considering the results from both the Categorization tasks and Encoding details tasks of the first experiment and similar results from the validation experiment, we emphasize the change to and from third trials of the main experiment in the fMRI results.
- Evaluation of the effect of categorization on subordinate details was followed by examination of the recall rates during the Encoding details tasks, and a significant “dipping” in recall rate was found prior to the assumed time point of categorization.
- For the experiment, participants were scanned while completing given tasks in a total of four sessions, alternating between experimental and control sessions. To evaluate whether attending superordinate for semantic categorization disrupts or enhances subordinate episodic details, we designed experimental sessions consisting of four sets of five trials that included an Encoding detail words task and a Categorization task in each trial, as well as a single Confirmation task after the fifth trial.
- Functional imaging was conducted on a 3T Siemens MAGNETOM Trio, A Tim System MRI scanner, and functional data was acquired using a gradient-echo planar pulse sequence (repetition time = 2000 msec, TE = 30 msec, 3 x 3 x 3 mm resolution, 33 axial slices tilted 30° from the AC-PC plane, no gap, interleaved collection).
- However, the effect of this grouping on subordinate levels of memories has not yet been well researched. Here, we examined whether semantic categorization of superordinates would enhance or disrupt its subordinate details by implementing a new paradigm that requires simultaneous encoding and retention of episodic details while engaging in categorization of superordinates. Employing RSA with searchlight in addition to conventional GLM analysis, we sought to examine not only the level of neural activation of a certain region, but also the pattern similarity with voxel-level sensitivity (Kriegeskorte, Goebel & Bandettini, 2006; Kriegeskorte, Mur & Bandettini, 2008).
- Comparisons of the third to fourth and fourth to fifth trials resulted in no important supra-threshold activation difference. Higher levels of neural activation in the areas of the semantic network found in the analysis implies usage of cognitive resources in the process of semantic categorization and supports our behavioral results in which categorization was postulated to be completed between the second and third trials. (see Fig 4 and Table 1)
- In control sessions, target words in a set were not semantically related to each other, nor were set words related to target words; however, detail words were presented in the same manner as in the experimental sessions. In the post-scan cued recall test outside of the scanner, set words from the experiment were presented in random order on a computer screen in red, and participants typed as many associated detail words as they could remember in a box provided underneath the target word on the screen, in a self-paced manner.
- Our study tested the effect of semantic categorization on subordinated detail using a new paradigm that engages encoding and retention of details with categorization of superordinate, and both behavioral and neural results identified disruption of episodic details prior to completion of categorization. In spite of the limitations in our experiment, the current results elucidate an unknown effect of categorization on subordinate details.
- Sequential increases and decreases in pattern consistencies with third trials showing the highest pattern consistencies was found even after removing the voxels responsible for statistical differences in GLM analysis. This result indicates that pattern similarities in the processing of semantic information convey different meanings than those of overall magnitude changes in neural activation.
- The detailed procedure for the experimental sessions is as follows: in the Encoding detail words task, a red-colored target word with two black-colored detail words beneath it appeared at the center of the screen and participants were asked, “Are the black words related to the red word?
- A set word presented in black in the Confirmation task was either associated with five target words or chosen randomly. The experimental runs were designed in a way that participants were to assume the five target words belonged to a single category, while the presenting order of the target words in a set was randomly chosen. A set word appeared after the fifth trial, which was either the “single” category that participants assumed the five target words were under or randomly chosen word, by the same chance (For a full list of stimuli, please refer to Appendix A).
- The functional images were realigned to correct for head movements, spatially normalized to the Montreal Neurological Institute (MNI) template provided with SPM8, then resampled into 3-mm cubes, followed by spatial smoothing with a 8-mm full-width, half-maximum isotropic Gaussian kernel. For GLM analyses, volumes were treated as temporally correlated time series and modeled by convolving canonical hemodynamic response function (HRF) and its temporal derivative, with delta function marking the onset of each trial.
- To verify the found effect, we also conducted a behavioral validation experiment that had manipulated difficulty of categorization. The new experiment was designed in a way that the point of categorization would be pushed forward along with the temporal position of the disruption on encoding of detail words. The results followed the design that both categorization completion and the time point of encoding disruption were delayed.
- The new findings in our study on the effect of categorization of superordinate upon episodic details highlight the loss of memory details, but it does not conflict with well-known previous studies as our research pertains to different levels of mnemonic structures involved in categorization. Unlike conventional categorization experiments dealt with memories of the information being categorized and discussed how categorization affects information at the same semantic level, we analyzed different level of information; subordinate memories.
- We analyzed the reaction times (RTs) and rated confidence in Categorization tasks, and RTs and recall rates from cued recalled test of Encoding details tasks. The time point of categorization completion was assumed when significant changes were simultaneously observed in RTs and confidence ratings of Categorization tasks, and the effect of categorization on subordinate details was evaluated by examining the recall rates of Encoding details tasks around the assumed time point.
- Comparisons of third to fourth and fourth to fifth trials did not show important supra-threshold differences in activation. These results support the findings from our behavioral study, in which categorization was possibly completed between second and third trials. In the Encoding details tasks, no significant difference in neural activation in each trial was expected, as every trial was supposed to be completed with a similar amount of effort.
- To corroborate our findings, the GLM analysis was repeated with a very liberal threshold (p < .1, uncorrected), identifying the voxels with significant differences in activation magnitudes.
- For the experiment, participants were scanned while completing given tasks in a total of four sessions, alternating between experimental and control sessions. To evaluate whether attending superordinate for semantic categorization disrupts or enhances subordinate episodic details, we designed experimental sessions consisting of four sets of five trials that included an Encoding detail words task and a Categorization task in each trial, as well as a single Confirmation task after the fifth trial. Prior to the scanning, participants repeatedly completed practice sessions which involved the same procedure to the experimental sessions, until they were familiarized with the tasks.
대상 데이터
- Nineteen undergraduate students (6 males, 13 females) participated in our study for either course credit or monetary compensation. Experimental tasks and procedures were identical to the fMRI experiment, except for the word stimuli; 9 target words and 18 detail words used in the experimental sessions were replaced (22.
데이터처리
- A beta estimate for each trial was collected voxel-by-voxel over the whole brain using a 3 x 3 x 3 voxel searchlight, and then the data was sorted according to the type of tasks involved. Inter-trial cross-correlation of the data was calculated for each subject and pair-wise contrasted at the group level with two-tailed t-tests. Resulting p-values of t-tests for each voxel were saved into a separate image for each contrast.
- The resulting hemodynamic functions were used as covariates in a general linear model along with a basis set of cosine functions used to high-pass filter the data and covariates representing session effects. Least-square parameter estimates of the best-fitting synthetic HRF for each condition of interest (averaged across scans) were used in pair-wise contrasts and were stored as separate images for each subject, then checked against the null hypothesis with one-tailed t-tests to determine whether effects of subjects were random at the group level. The single trial analysis (STA) which considers every trial of interest as a separate regressor, was applied with SPM8 for RSA for inter-trial comparisons of embedded neural pattern consistencies in order to reveal additional meaningful information that was unobservable by collapsing data to a mean (Pernet, Sajda & Rousselet, 2011).
성능/효과
- , 1999; Manns, Clark, & Squire, 2002; Rajah & McIntosh, 2005; Wagner, Bunge & Badre, 2004) for the fMRI experiment yielded results that showed similarities to the behavioral results, and the level of effort required for categorization was expected to be lessened, trial-by-trial, until the moment of successful categorization. As expected, we observed significantly lower neural activation of the bilateral dlPFC and precunei, rSFG, rSTG, rMTG, rITG, and caudate nuclei in third trials compared to second trials, and no significant change in activation was observed after the third trials. Comparisons of third to fourth and fourth to fifth trials did not show important supra-threshold differences in activation.
- 008. Behavioral changes from one trial to the next were similar to the direction of results from the main experiment, but a significant decrease of RT and an increase of confidence were found in every sequential comparison of trials. Based on these new results, we posit that categorization was completed in the last trial, due to increased difficulty of the Categorization task compared to our prior experiment.
- Higher overall activations were found in the first trials compared to second trials but were disregarded as those were possibly caused by a novelty effect of the task or stimuli rather than actual neural activity of our interest. Comparison of first to the second trials showed significantly higher activation in the left superior frontal gyrus (lSFG; BA 8/9/10), bilateral inferior frontal gyri (IFG; BA 9/47) and medial frontal gyrus (mFG; BA 6/10/11/25) of the dorsolateral prefrontal cortex (dlPFC), bilateral superior temporal gyri (STG; BA 22/38/39/42) and middle temporal gyrus (mTG; BA 19/21/38/39), bilateral precunei (PrC; BA 7/31) and parahippocampal gyrus (PHG; BA 19/35/36). The right medial frontal gyrus (rmFG; BA 10) was found to be more activated in the second trials compared to third trials, while the comparison in the opposite direction did not show any significant difference in our ROIs.
- Representational similarity analysis (RSA) with searchlight in addition to conventional general linear model (GLM) analysis was applied to achieve more sensitive measurements, as RSA sensitively examines consistency in the patterns of neural activity rather than the intensity of the activation (Kriegeskorte, Goebel & Bandettini, 2006; Kriegeskorte, Mur & Bandettini, 2008). Hypothesizing that progression of categorization toward completion would show stronger pattern similarity, we sought to identify distinguishable pattern consistencies regarding the process with RSA, while GLM analysis was used to investigate the level of effort involved in processing episodic details, which was not expected to differ throughout the experiment.
- Confidence ratings and RTs of Encoding details tasks were not considered for analysis, as both are more associated with episodic memory formed in Encoding details tasks rather than sequential categorization. In Categorization tasks, pair-wise t-tests between trials in sequential order revealed that RT significantly decreased from second trials (M = 2142.54 ms) to third trials (M = 1727.38 ms), t(17) = 3.52, p = .003, while no significant difference was found between third and fourth (M = 1567.38 ms) trials, t(17) = 1.51, p = .15, or fourth and fifth trials (M = 1570.42 ms), t(17) = -0.04, p = .97. In opposition to the observed decrease in RT, confidence ratings increased from second (M = 3.
- This result indicates that pattern similarities in the processing of semantic information convey different meanings than those of overall magnitude changes in neural activation. In support of the RSA results, these new results possibly indicate that neural patterns in the categorization process change based on its progress toward completion and are reliably distinguished from neural activation strength, although not completely independent of the fluctuation in neural firing intensities.
- Considering the differences found in both RTs and confidence ratings, we postulate that categorization is successfully established while progressing from second to third trials. In the Encoding details tasks, RT only decreased significantly between the first (M = 2305.77 ms) and second trials (M = 2102.06 ms), t(17) = 2.49, p = .02, and did not show a significant decrease from second to third (M = 2072.13 ms), t(17) = 0.36, p = .72, third to fourth (M = 2085.06 ms), t(17) = -0.12, p = .90, or fourth to fifth trials (M = 1928.83 ms), t(17) = 1.68, p = .11.
- Second, even though we tried to minimize the sensitivity problem by regulating the number of voxels included in a searchlight to a relatively conservative number, RSA with searchlight can be overly sensitive due to the fact that each voxel holds the mean correlation value of included trials’ activations.
- This could also be true for the validation experiment, but a chance to explore such a tendency was not available in our new experiment, since categorization was found to be established in the last trial. Second, the result that subsequent cued recall rates of detail words in the validation experiment differed only in sequential comparison of the fourth and fifth trials could be seen as categorization is completed before participants reached the fifth trials. We call the significantly low recall rate prior to the point of categorization a “dip” not only due to the peaking of recall rate after the “dip”, but also because only the fourth trials were found to be significantly different from fifth trials compared to all other trials.
- The analysis concluded as we expected, except that the overall activation pattern was found be higher on first trials compared to second trials which was disregarded as a novelty effect of the task or stimuli.
- The recall rates “dipped” prior to third trials, as a marginally significant decrease in recall rate from the first (M = .36) to second trials (M = .30) was found, t(17) = 1.87, p = .08, while recall rate significantly increased from the second to third trials (M = .40), t(17) = -3.07, p = .007.
- In Categorization tasks, the experimental procedures prompted participants to actively engage in the categorization process, and the level of effort required was anticipated to decrease with trials until the point of categorization completion. When comparing the neural activities of Categorization tasks involved in second and third trials, higher activations were observed in the second trial in the bilateral dorsolateral prefrontal cortices (dlPFC) including the bilateral inferior frontal gyri (IFG; BA 45), middle frontal gyrus (MFG; BA 6/8/9/46), right superior frontal gyrus (rSFG; BA 6), and in right superior temporal gyrus (rSTG; BA 22), middle temporal gyrus (rMTG; BA 21/22), inferior temporal gyrus (rITG; BA 20), and bilateral precunei (PrC; BA 7/19/31) and caudate nuclei, while no significant hyperactivity was found in the opposite direction of the comparison. Comparisons of the third to fourth and fourth to fifth trials resulted in no important supra-threshold activation difference.
후속연구
- Although our hypothesis was supported by experimental results, further investigation on factors involved in the resulting phenomenon were needed to make sure that the results found in specific time periods was actually occurred due to the factors we suspected. We conducted a behavioral validation experiment by manipulating the difficulties in the Categorization tasks, moving forward the temporal position of categorization completion.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.