Abstract

Memories of past episodes frequently come to mind incidentally, without directed search. It has remained unclear how incidental retrieval processes are initiated in the brain. Here we used fMRI and ERP recordings to find brain activity that specifically correlates with incidental retrieval, as compared to intentional retrieval. Intentional retrieval was associated with increased activation in dorsolateral prefrontal cortex. By contrast, incidental retrieval was associated with a reduced fMRI signal in posterior brain regions, including extrastriate and parahippocampal cortex, and a modulation of a posterior ERP component 170 msec after the onset of visual retrieval cues. Successful retrieval under both intentional and incidental conditions was associated with increased activation in the hippocampus, precuneus, and ventrolateral prefrontal cortex, as well as increased amplitude of the P600 ERP component. These results demonstrate how early bottom–up signals from posterior cortex can lead to reactivation of episodic memories in the absence of strategic retrieval attempts.

INTRODUCTION

Episodic memory is a system for storing and retrieving information about personally experienced events. Retrieving information from episodic memory occurs via interactions between a retrieval cue and a stored memory trace (Tulving, 1983). There is much evidence for the importance of prefrontal cognitive control processes in the intentional retrieval of episodic memories, and it has been argued that a prefrontally supported “retrieval mode” is a prerequisite for retrieval of past events to occur (Lepage, Ghaffar, Nyberg, & Tulving, 2000; Wheeler, Stuss, & Tulving, 1997). Although the specific contribution of prefrontal cortex in supporting retrieval remains unclear, its suggested roles include specifying the retrieval cues and search parameters, monitoring the outcome of retrieval attempts, and processing the retrieved information (Fletcher & Henson, 2001). The dorsolateral part of prefrontal cortex has been demonstrated to provide top–down input to the medial temporal lobe in the service of retrieval (Tomita, Ohbayashi, Nakahara, Hasegawa, & Miyashita, 1999). On the other hand, memories of past episodes may also come to mind incidentally, without any intentional strategic search (Berntsen, 1996; Richardson-Klavehn & Gardiner, 1995). Medial temporal and parietal brain activity associated with retrieval success does not seem to be affected by lack of retrieval intention (Hall, Gjedde, & Kupers, 2008; Curran, 1999; Rugg, Fletcher, Frith, Frackowiak, & Dolan, 1997), but there is evidence that prefrontal recruitment is not salient during incidental retrieval (Hall et al., 2008; Rugg et al., 1997). The failure to find robust prefrontal activation during incidental retrieval raises the question of how such instances of retrieval are initiated in the brain. This question has received relatively little attention in memory research, and does not yet have a comprehensive explanation. Although behavioral studies suggest that incidental retrieval, too, is associated with a retrieval cue, most often a salient stimulus in the surrounding environment (Berntsen, 1996), it is unclear how such stimuli can engage the episodic memory system. Single-cell recordings in monkeys have suggested that bottom–up signals from temporal cortex neurons can elicit “automatic” retrieval, independent of top–down support, in a visual paired associate task (Osada, Adachi, Kimura, & Miyashita, 2008; Miyashita, 2004; Naya, Yoshida, & Miyashita, 2001). These studies raise the interesting possibility of a “low route” to memory such that bottom–up signals indicating stimulus oldness arise during the processing of potential retrieval cues and lead to episodic retrieval without top–down input from prefrontal cortex. It has been argued that parahippocampal and visual association cortices are capable of rapidly and automatically computing visual stimulus oldness without top–down input from either the hippocampus or prefrontal cortex (Brown & Aggleton, 2001), providing a potential source for such bottom–up input. The possibility of episodic retrieval processes initiated automatically by the hippocampus as a reaction to suitable input, regardless of whether it arises as a result of bottom–up or top–down processes, is considered in some theoretical models of episodic memory, most notably by Moscovitch (1992) (see also Conway & Pleydell-Pearce, 2000).

This study was performed to identify the spatial and temporal patterns of brain activity specifically associated with incidental episodic retrieval.1 We report the findings from two experiments using fMRI (Experiment 1) and ERPs (Experiment 2). We examined incidental retrieval of episodic memories in the context of a novel memory-categorization dual task. By varying the intention to retrieve within this task, we could isolate the brain activity contributing to retrieval only in the incidental condition (the hypothesized bottom–up input). We predicted that dorsolateral and anterior prefrontal cortex, associated with strategic top–down control during a retrieval task (Fletcher & Henson, 2001), would be activated only during intentional retrieval, independently of retrieval success. By contrast, incidental retrieval should be preceded by distinct neural signals indicating stimulus oldness, arising from parahippocampal and/or visual association areas and being expressed as differences in ERP amplitude during the early stages of processing. Based on previous findings (Hall et al., 2008; Curran, 1999; Rugg et al., 1997), we expected that the neural activation associated with retrieval success, including the BOLD signal amplitude in the hippocampus and precuneus, as well as the P600 ERP component amplitude, should not be affected by retrieval intention.

EXPERIMENT 1

Subjects

In the fMRI data collection, the subjects were 16 healthy young right-handed adults (mean age = 24.1 ± 4.5 years; 10 women) with normal or corrected-to-normal vision and normal hearing. Two of the subjects were excluded as behavioral outliers due to too few incidentally retrieved items (<9 items), resulting in 14 subjects. The subjects gave written informed consent prior to participation. The study was approved by the ethics committee at the University Hospital of Northern Sweden.

Behavioral Paradigm

A schematic overview of the experimental protocol used in this study is presented in Figure 1. Approximately 24 hr before the fMRI and ERP data collection, the subjects encoded 40 sounds (naturalistic environmental sounds) and 40 pictures (color photographs) representing common, easily identifiable objects (tools, animals, music instruments, household items, etc.). Varying the encoding modality was involved to allow for an additional, independent analysis of the resulting data; the encoding modality was not used as a variable of interest in the present study. The sounds and pictures were presented in separate, unrelated lists. Each picture was presented for 2500 msec on a computer screen; the sounds varied in duration between 2000 and 4000 msec. The sounds were presented via headphones with individually adjusted volume. As behavioral pilot testing had revealed that the sounds were more difficult to retrieve, the sound list was presented twice, with the order of the sounds randomly intermixed between the two presentations. The subjects were instructed to memorize the items and were told that a memory test would be administered on the following day (no information on the testing procedure was given). During the retrieval tasks (when the imaging data were collected), the subjects were visually presented with concrete and abstract words (n = 240; presentation duration 2500 msec; word order pseudorandomized). In the intentional retrieval condition (consisting of 120 words), the subjects were instructed to identify the words which matched the objects that were previously encoded in sound or picture format (without indicating the modality); in the incidental retrieval condition (consisting of 120 words), they were instructed to categorize the words as concrete or abstract. During instructions, the two conditions were referred to as “the memory task” and “the categorization task.” In the intentional condition, the subjects were additionally instructed to categorize each word, whereas in the incidental condition, they were instructed to report in case a word happened to spontaneously elicit a memory of a previously encoded object (i.e., during both intentionally and incidentally retrieved words two responses were given, whereas during categorization of new concrete words only one response was given in both conditions). It was emphasized in the instructions that the subjects should entirely concentrate on retrieval during the intentional condition, and on categorization during the incidental condition, and attempt to maximize their performance. The proportion of new concrete, matching concrete, and abstract words was identical in both conditions (40 instances of each category). The order of the tasks was fixed across subjects, with the incidental condition following the intentional condition in order to minimize retrieval attempts during the incidental condition. The rationale for fixing the order of the tasks was derived from qualitative interviews performed during pilot testing when the experimental protocol was being developed. The pilot subjects (n = 5, no overlap with the subjects from whom the physiological data were gathered) reported engaging in more deliberate retrieval attempts (despite explicit instructions to the contrary) during the incidental retrieval task when it was not preceded by intentional retrieval task, whereas placing the incidental retrieval after an intentional retrieval task led to few reports of the use of deliberate strategies. The low proportion of words that matched the studied items, combined with a shift in stimulus modality between study and test (Habib & Nyberg, 1997), as well as a long retention interval, served to additionally minimize the influences of intentional strategies in the incidental condition. This experimental protocol allowed us to compare brain activity corresponding to the trials which were equivalent in all other respects apart from retrieval processing: retrieved (successfully remembered matching concrete words) and new concrete (novel concrete words). By varying the intention to retrieve, our protocol corresponded to a 2 × 2 factorial design with the factors task (intentional/incidental) and item (retrieved/new concrete). The subjects responded with a button press, using three fingers of their right hand (index, middle, ring = concrete, abstract, remember). The order of the categorization and remember responses was not fixed, but the subjects were instructed to report the memories immediately when they occurred, whether before or after the categorization response. In-between the words, a fixation cross was displayed on the screen (duration 2000 msec). Throughout the recordings, there were regularly spaced resting intervals after every 10 words (duration = 20 sec), during which a fixation circle was displayed. E-Prime software (1.1, Psychology Software Tools, Inc., Pittsburgh, PA) was used to control the presentation of the stimuli and to record the responses.

Figure 1. 

The experimental protocol consisted of an encoding phase, where the participants memorized 40 natural sounds and 40 color photographs of common, easily recognizable concrete objects, which was followed about 24 hr later by two separate retrieval tasks under intentional and incidental conditions. The imaging (fMRI or EEG) and behavioral data were collected only during the retrieval phase. Under the intentional condition, the participants were instructed to concentrate on retrieval, but additionally, categorize each item. Under the incidental condition, they were instructed to concentrate on categorization, but report incidentally retrieved items. Intentional condition always occurred before incidental condition to minimize retrieval attempts in the latter condition. The participants responded with three fingers of their right hand in both tasks to indicate concrete (C), abstract (A), and remember (R) responses (see Behavioral Paradigm section for additional details).

Figure 1. 

The experimental protocol consisted of an encoding phase, where the participants memorized 40 natural sounds and 40 color photographs of common, easily recognizable concrete objects, which was followed about 24 hr later by two separate retrieval tasks under intentional and incidental conditions. The imaging (fMRI or EEG) and behavioral data were collected only during the retrieval phase. Under the intentional condition, the participants were instructed to concentrate on retrieval, but additionally, categorize each item. Under the incidental condition, they were instructed to concentrate on categorization, but report incidentally retrieved items. Intentional condition always occurred before incidental condition to minimize retrieval attempts in the latter condition. The participants responded with three fingers of their right hand in both tasks to indicate concrete (C), abstract (A), and remember (R) responses (see Behavioral Paradigm section for additional details).

Acquisition and Preprocessing of the fMRI Data

The fMRI data were collected using a 3-T Philips Achieva scanner at the University Hospital of Northern Sweden. Functional T2*-weighted images were acquired with a single-shot gradient-echo sequence (TR = 1512 msec, TE = 30 msec, FOV = 22 cm). After the functional image acquisition, the structural high-resolution T1 images (170 slices, 1 mm thickness, FOV = 24 cm) were collected. The images were analyzed using SPM5 software (Wellcome Trust Centre for Neuroimaging, London, UK). An in-house-developed MATLAB-based software was used for batching the preprocessing and first-order analyses as well as for visualizing the results. The functional images were realigned and unwarped, corrected for within-volume time delay due to slice-wise acquisition, normalized to the Montreal Neurological Institute T1-weighted template, and smoothed with a Gaussian kernel (full width at half maximum = 8 mm).

Statistical Analysis of the fMRI Data

A general linear model was set up, including eight regressors, modeling the effects corresponding to all word types (retrieved, failed, new concrete, and abstract in both intentional and incidental tasks). The regressors consisted of delta functions placed at the stimulus onset, convolved with a canonical hemodynamic response function, allowing to model the transient portion of the BOLD signal changes. The group data were analyzed in a two-step procedure. First, for each subject, the model was estimated in each voxel, resulting in eight beta images corresponding to the eight regressors. For each of the four conditions of interest (intentional retrieved, intentional new concrete, incidental retrieved, incidental new concrete), t contrasts were performed ([1 0 0 0], [0 1 0 0], etc). Then the four t-contrast images from each subject were entered into a 14 × 4 (Subject × Condition) repeated measures ANOVA (using SPM5 software). Following this, the condition effects were explored using F-contrasts, testing for main effect of task, main effect of item, and the interaction between task and item. The contrast representing the main effect of the task was thresholded at p < .0001, uncorrected for multiple comparisons (cluster extent threshold 5 voxels). For the contrast representing main effect of the item, the threshold was increased to p < .01, with correction for family-wise error to allow better separation of the regional clusters (cluster extent threshold 5 voxels). As we specifically wanted to examine semi-disordinal interaction patterns in order to find brain regions where the BOLD signal would not be modulated by retrieval within the intentional condition but would instead be modulated within the incidental condition (and vice versa), we performed a two-step interaction analysis. First, we used t contrasts to find regions where incidental retrieval caused increased/decreased activation compared to the average of other three conditions ([−1 −1 3 −1] and [1 1 −3 1], respectively); we then inclusively masked each of these contrasts (thresholded at p < .0001, uncorrected) with the Task × Item interaction contrast from the repeated measures ANOVA (thresholded at p < .01, uncorrected). The same procedure was used to find the brain regions modulated within intentional, but not incidental, condition.

Region-of-interest Analysis in the Hippocampus

Considering the extensive evidence from lesion and imaging studies for a role of the hippocampus in episodic memory retrieval (Eldridge, Knowlton, Furmanski, Bookheimer, & Engel, 2000; Schacter & Wagner, 1999; Nyberg, McIntosh, Houle, Nilsson, & Tulving, 1996; Cohen & Squire, 1981; Milner, Corkin, & Teuber, 1968), we performed a region-of-interest analysis of the BOLD signal in the posterior hippocampus bilaterally to examine in detail the effect of intention on retrieval-related activity in the hippocampus. Each subject's high-resolution T1 image was coregistered to the mean functional image and normalized using the functional data normalization parameters. In the coregistered and normalized structural image, regions of interest (2 mm sphere) within the hippocampus were defined bilaterally in the caudal portion of the body. Next, using the MarsBaR toolbox (http://marsbar.sourceforge.net), for each subject, the mean time course for the voxels falling within each ROI was calculated, and the previously defined general linear model (see above) was fitted to that summary time course. Parameter estimates corresponding to the four conditions of interest (retrieved and new concrete words within intentional and incidental condition) were subjected to a 2 (laterality)× 2 (task) × 2 (item) repeated measures ANOVA (using SPSS Statistics 17.0 software).

Results

Behavioral Results

The behavioral data confirmed that the subjects solved the task differently in the two conditions. They retrieved more items in the intentional than in the incidental condition [intentional: 61.6%; incidental: 47.4%; t(13) = 4.13, p < .005]. The “remember” response was given as the first response (compared to the categorization response for the same item) more often in the intentional condition [intentional: 51%; incidental: 6%; t(13) = 4.77, p < .005], and retrieval was, correspondingly, faster [intentional:1594 msec; incidental: 1780 msec; t(13) = 3.8, p < .005].

Additional evidence came from the response times to the new concrete words (which did not match any of the encoded items). In the intentional task, the new concrete words were categorized significantly slower than the abstract words [1485 msec vs. 1210 msec; t(13) = 6.25, p < .005]. Such a difference was not seen in the incidental condition (new concrete: 1194 msec; abstract: 1101 msec), indicating that the subjects evaluated the concrete items in the intentional task to decide whether they matched the previously encoded items, whereas such evaluation was not performed in the incidental condition. Taken together, the results suggest that a directed search guided memory retrieval only in the intentional condition.

fMRI Results

The fMRI data were analyzed with a 2 (task: intentional vs. incidental) × 2 (item: retrieved vs. new concrete) repeated measures ANOVA. First, we present the brain regions showing a significant main effect of task. We found that the BOLD signal in right dorsolateral prefrontal cortex was selectively enhanced in the intentional condition (Figure 2A). In addition, the intention to retrieve increased the BOLD signal in left anterior prefrontal cortex, parietal cortex, and cerebellum (see Table 1). The fact that these regional responses were seen in the intentional condition, regardless of whether an item was successfully retrieved or, instead, correctly categorized, suggests that these brain regions are related to the strategic aspects of retrieval rather than accessing a stored memory per se (Nyberg et al., 1995). The finding of dorsolateral and anterior prefrontal activation only in the intentional condition converges with the behavioral data in suggesting a fundamental distinction between the intentional and incidental retrieval conditions. There were no regions showing a main effect of task in the opposite direction (incidental > intentional). Next, we examined the regions showing a significant main effect of item. Successful retrieval, regardless of intentionality, was expressed as an increased BOLD signal in inferior prefrontal cortex bilaterally, ACC, precuneus, lateral parietal cortex, and in the brain stem (Figure 2B and Table 2). There were no regions where the main effect of item showed the opposite direction (new > retrieved).

Figure 2. 

The retrieval intention and retrieval success were expressed in distinct brain regions. (A) Retrieval intention activated dorsolateral prefrontal cortex (projected on a T1-weighted anatomical template; MNI coordinates of peak voxel in mm: x, y, z = 44, 42, 26; Z = 4.14; p < .0001, uncorrected). Bar graph, showing the percent of BOLD signal change compared to the mean of session, shows that dorsolateral prefrontal cortex was activated under the intentional condition, regardless of whether the trial contained an old item which was retrieved (R) or a new concrete item which was categorized (C). By contrast, dorsolateral prefrontal cortex was not significantly activated under the incidental condition. Lines indicate the standard error of the mean. (B) Successful retrieval, regardless of intentionality, activated brain regions in (1) right vlPFC (coordinates of peak voxel in mm: x, y, z = 34, 20, −10; Z = 6.53; p < .01, corrected for family-wise error), (2) ACC (x, y, z = −6, 26, 36; Z = 6.45; p < .01, corrected for family-wise error), and (3) medial parietal cortex (x, y, z = −10, −72, 32; Z = 5.54; p < .01, corrected for family-wise error). Line graphs, showing the percent of BOLD signal change compared to the mean of session, shows that successful retrieval engaged these regions similarly, despite the difference in the retrieval intentionality between the conditions. Crosslines indicate the standard error of the mean. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words.

Figure 2. 

The retrieval intention and retrieval success were expressed in distinct brain regions. (A) Retrieval intention activated dorsolateral prefrontal cortex (projected on a T1-weighted anatomical template; MNI coordinates of peak voxel in mm: x, y, z = 44, 42, 26; Z = 4.14; p < .0001, uncorrected). Bar graph, showing the percent of BOLD signal change compared to the mean of session, shows that dorsolateral prefrontal cortex was activated under the intentional condition, regardless of whether the trial contained an old item which was retrieved (R) or a new concrete item which was categorized (C). By contrast, dorsolateral prefrontal cortex was not significantly activated under the incidental condition. Lines indicate the standard error of the mean. (B) Successful retrieval, regardless of intentionality, activated brain regions in (1) right vlPFC (coordinates of peak voxel in mm: x, y, z = 34, 20, −10; Z = 6.53; p < .01, corrected for family-wise error), (2) ACC (x, y, z = −6, 26, 36; Z = 6.45; p < .01, corrected for family-wise error), and (3) medial parietal cortex (x, y, z = −10, −72, 32; Z = 5.54; p < .01, corrected for family-wise error). Line graphs, showing the percent of BOLD signal change compared to the mean of session, shows that successful retrieval engaged these regions similarly, despite the difference in the retrieval intentionality between the conditions. Crosslines indicate the standard error of the mean. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words.

Table 1. 

Brain Regions Showing Significant Main Effect of Task (Intentional > Incidental; p < .0001, Not Corrected for Multiple Comparisons; Cluster Extent Threshold 5 Voxels)

Region
BA
x
y
z
Z
k
Frontal 
Middle frontal gyrus 26 50 5.81 747 
Superior frontal gyrus −26 70 4.26 105 
Middle frontal gyrus 46 44 42 26 4.14 21 
Superior frontal gyrus 10 −32 58 22 3.97 
 
Parietal 
Parietal 24 −50 56 5.34 622 
Postcentral gyrus −12 −62 70 5.06 1179 
Postcentral gyrus 40 50 −36 50 4.27 74 
Paracentral lobule 18 −40 54 3.98 
 
Temporal 
Fusiform gyrus 20 −46 −24 −18 4.74 60 
Inferior temporal gyrus 20 60 −54 −16 3.86 
Inferior temporal gyrus 37 52 −70 −6 3.82 
 
Other 
Hypothalamus  −2 −10 4.02 14 
Cerebellum  −4 −78 −18 5.28 928 
Cerebellum  −6 −80 −42 4.16 31 
Cerebellum  32 −68 −24 4.04 24 
Cerebellum  24 −36 −32 3.98 
Cerebellum  28 −54 −34 3.96 17 
Region
BA
x
y
z
Z
k
Frontal 
Middle frontal gyrus 26 50 5.81 747 
Superior frontal gyrus −26 70 4.26 105 
Middle frontal gyrus 46 44 42 26 4.14 21 
Superior frontal gyrus 10 −32 58 22 3.97 
 
Parietal 
Parietal 24 −50 56 5.34 622 
Postcentral gyrus −12 −62 70 5.06 1179 
Postcentral gyrus 40 50 −36 50 4.27 74 
Paracentral lobule 18 −40 54 3.98 
 
Temporal 
Fusiform gyrus 20 −46 −24 −18 4.74 60 
Inferior temporal gyrus 20 60 −54 −16 3.86 
Inferior temporal gyrus 37 52 −70 −6 3.82 
 
Other 
Hypothalamus  −2 −10 4.02 14 
Cerebellum  −4 −78 −18 5.28 928 
Cerebellum  −6 −80 −42 4.16 31 
Cerebellum  32 −68 −24 4.04 24 
Cerebellum  24 −36 −32 3.98 
Cerebellum  28 −54 −34 3.96 17 

BA = Brodmann’s area; x, y, z = coordinates of peak voxel in MNI space; Z = Z-score of peak voxel; k = number of activated voxels.

Table 2. 

Brain Regions Showing Significant Main Effect of Item (Retrieved > New Concrete, p < .01, Corrected for Family-wise Error; Cluster Extent Threshold 5 Voxels)

Region
BA
x
y
z
Z
k
Frontal 
Inferior frontal gyrus 47 −32 22 −4 7.14 622 
Inferior frontal gyrus 47 34 20 −10 6.53 388 
Anterior cingulate gyrus 32 −6 26 36 6.45 377 
 
Parietal 
Postcentral gyrus −38 −26 54 6.88 702 
Cuneus −10 −72 32 5.54 25 
 
Other 
Brainstem  −4 −24 −2 6.81 746 
Lentiform nucleus  −18 −2 6.04 152 
Region
BA
x
y
z
Z
k
Frontal 
Inferior frontal gyrus 47 −32 22 −4 7.14 622 
Inferior frontal gyrus 47 34 20 −10 6.53 388 
Anterior cingulate gyrus 32 −6 26 36 6.45 377 
 
Parietal 
Postcentral gyrus −38 −26 54 6.88 702 
Cuneus −10 −72 32 5.54 25 
 
Other 
Brainstem  −4 −24 −2 6.81 746 
Lentiform nucleus  −18 −2 6.04 152 

BA = Brodmann’s area; x, y, z = coordinates of peak voxel in MNI space; Z = Z-score of peak voxel; k = number of activated voxels.

For the region-of-interest analysis in the hippocampus, a 2 (laterality) × 2 (task) × 2 (item) repeated measures ANOVA showed a significant main effect of item [retrieved > new concrete; F(1, 13) = 13.63, p = .003]; no other effects were significant. Thus, successful retrieval increased the strength of the BOLD signal in the hippocampus regardless of intentionality (Figure 3A).

Figure 3. 

Intentional and incidental retrieval exhibited similarities in the hippocampus activation, but incidental retrieval elicited selective response in the posterior brain regions, including extrastriate cortex. (A) Intentionally and incidentally retrieved memories both involve the hippocampus, further confirming that the actual reactivation of a memory was mediated by similar processes independently of retrieval intentionality. The line graph illustrates the regional response amplitude from the posterior hippocampus (pooled across left and right hemisphere regions of interest). The crosshair projected on the T1-weighted anatomical template illustrates the approximate location of the region of interest in the hippocampus, selected individually for each participant. (B) The relative reduction of the BOLD signal intensity in left extrastriate cortex (Brodmann’s area 19, MNI x, y, z = −32, −68, 0; Z = 4.15; p < .0001, uncorrected) was selectively associated with incidental retrieval. Such reduction was not observed under the intentional condition. The bar graph, showing the percent of BOLD signal change compared to the mean of session, highlights that under the incidental condition the items which the subjects failed to retrieve (gray bar) did not significantly differ from the new concrete items in terms of the neural response. This indicates that the BOLD signal reduction constituted a signal triggering retrieval under the incidental condition. Lines indicate the standard error of the mean. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words; F = failed words.

Figure 3. 

Intentional and incidental retrieval exhibited similarities in the hippocampus activation, but incidental retrieval elicited selective response in the posterior brain regions, including extrastriate cortex. (A) Intentionally and incidentally retrieved memories both involve the hippocampus, further confirming that the actual reactivation of a memory was mediated by similar processes independently of retrieval intentionality. The line graph illustrates the regional response amplitude from the posterior hippocampus (pooled across left and right hemisphere regions of interest). The crosshair projected on the T1-weighted anatomical template illustrates the approximate location of the region of interest in the hippocampus, selected individually for each participant. (B) The relative reduction of the BOLD signal intensity in left extrastriate cortex (Brodmann’s area 19, MNI x, y, z = −32, −68, 0; Z = 4.15; p < .0001, uncorrected) was selectively associated with incidental retrieval. Such reduction was not observed under the intentional condition. The bar graph, showing the percent of BOLD signal change compared to the mean of session, highlights that under the incidental condition the items which the subjects failed to retrieve (gray bar) did not significantly differ from the new concrete items in terms of the neural response. This indicates that the BOLD signal reduction constituted a signal triggering retrieval under the incidental condition. Lines indicate the standard error of the mean. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words; F = failed words.

Next, to identify the neural correlates of the processes that trigger incidental retrieval, we tested for regions which selectively responded to retrieval (compared to new concrete items) in the incidental condition but did not differentiate between retrieved and new concrete items in the intentional condition (see Statistical Analysis section for details). A selective activation increase was found in posterior cingulate cortex (BA 23, x, y, z = 4, −18, 34; Z = 4.05; 11 voxels). Selective activation decreases were found in mainly posterior cortical regions, including bilateral superior temporal gyrus, precentral lobule, left parahippocampal cortex (BA 19/37) and left extrastriate cortex (BA 19; Figure 3B, Table 3). Post hoc t tests performed separately in each of the clusters confirmed that in none of these regions was the BOLD signal modulated by item type (retrieved vs. new concrete) in the intentional condition (p > .1). The results of this comparison could have been influenced by insufficient power to detect moderate-sized differences during the intentional condition. Examining the effect sizes showed that although small to moderate-sized differences could be observed (d = 0.04 to d = 0.49), these were not reliable, with 95% confidence intervals (CI) of the effect size encompassing zero in all instances. By contrast, the effect sizes of the corresponding comparison in the incidental condition were large (d = 0.8 to d = 1.6) and reliable (95% CI spanning above zero).

Table 3. 

Brain Regions Showing Selective Decrease of Activation for Incidentally Retrieved Items [t Contrast Testing for Decrease Compared to the Average of Other Regressors (p < .0001) Masked Inclusively with F-contrast Testing for Interaction between Task and Item (p < .01)]

Region
BA
x
y
z
Z
k
Parietal 
Paracentral lobule 5/7 18 −42 56 5.20 236 
Postcentral gyrus 30 −24 48 4.22 19 
 
Temporal 
Parahippocampal cortex 19/37 −36 −36 −2 4.80 39 
Superior temporal gyrus 22 42 −24 4.73 33 
Superior temporal gyrus 22 −58 −36 4.43 44 
Insula 13 −42 −14 20 4.37 
 
Occipital 
Lingual gyrus 19 −32 −68 4.15 25 
Region
BA
x
y
z
Z
k
Parietal 
Paracentral lobule 5/7 18 −42 56 5.20 236 
Postcentral gyrus 30 −24 48 4.22 19 
 
Temporal 
Parahippocampal cortex 19/37 −36 −36 −2 4.80 39 
Superior temporal gyrus 22 42 −24 4.73 33 
Superior temporal gyrus 22 −58 −36 4.43 44 
Insula 13 −42 −14 20 4.37 
 
Occipital 
Lingual gyrus 19 −32 −68 4.15 25 

BA = Brodmann’s area; x, y, z = coordinates of peak voxel in MNI space; Z = Z-score of peak voxel; k = number of activated voxels.

We performed the same analysis for the items that were retrieved intentionally to allow comparison with the selective activation pattern predicting incidental retrieval. This analysis yielded a distinct pattern of neural activity. Selective activation increases were found in spatially restricted portions of superior parietal regions (precuneus and lateral parietal cortex), right anterior inferior parietal cortex, and right perirhinal cortex (Table 4). Selective activation decrease was seen in the primary visual cortex (BA 17, x, y, z = 20, −96, −6; Z = 4.68; k = 55). Similarly as above, we performed post hoc t tests to confirm that no modulation of the BOLD signal occurred in these regions during incidental condition. Apart from moderate effects in the postcentral gyrus [t(13) = 3.1, p < .05; d = 0.4] and the claustrum [t(13) = 2.5, p < .05; d = 0.6], no clusters showed significant modulation of activity during incidental condition, whereas the effect sizes of the modulation during intentional condition were moderate to large in all clusters (d = 0.5 to d = 1.3), with 95% CI spanning above zero. This shows that the areas which were selectively associated with incidental retrieval were distinct from the regions contributing to intentional retrieval.

Table 4. 

Brain Regions Showing Selective Increase of Activation for Intentionally Retrieved Items [t Contrast Testing for Increase Compared to the Average of Other Regressors (p < .0001) Masked Inclusively with F-contrast Testing for Interaction between Task and Item (p < .01)]

Region
BA
x
y
z
Z
k
Postcentral gyrus −12 −54 68 5.16 24 
Superior parietal lobule 16 −72 60 5.14 15 
Middle frontal gyrus 22 50 4.78 
Postcentral gyrus 58 −26 38 4.60 93 
Parahippocampal gyrus 35 28 −18 −22 4.34 19 
Postcentral gyrus 30 −46 70 4.26 18 
Superior frontal gyrus 22 −10 76 4.17 
Claustrum 36 −24 4.08 
Region
BA
x
y
z
Z
k
Postcentral gyrus −12 −54 68 5.16 24 
Superior parietal lobule 16 −72 60 5.14 15 
Middle frontal gyrus 22 50 4.78 
Postcentral gyrus 58 −26 38 4.60 93 
Parahippocampal gyrus 35 28 −18 −22 4.34 19 
Postcentral gyrus 30 −46 70 4.26 18 
Superior frontal gyrus 22 −10 76 4.17 
Claustrum 36 −24 4.08 

BA = Brodmann’s area; x, y, z = coordinates of peak voxel in MNI space; Z = Z-score of peak voxel; k = number of activated voxels.

A reduction of the BOLD signal during the second compared to the first presentation of a stimulus has been related to priming, an implicit (nonconscious) form of memory which is thought to operate independently of explicit (conscious) memory (Henson, 2003). The extrastriate regions, where we found an activity reduction in the incidental condition in response to the word cues that matched previously studied sounds or pictures, have been implicated in visual, auditory, and cross-modal priming (Rugg et al., 1998; Wiggs & Martin, 1998), and have been suggested to mediate amodal conceptual processes (Badgaiyan, Schacter, & Alpert, 1999). Similarly, the left superior temporal gyrus has been associated with cross-modal priming (Schacter, Badgaiyan, & Alpert, 1999). Consequently, one explanation for the observed reduction in the BOLD signal amplitude for the incidentally retrieved items is that it is unrelated to incidental episodic retrieval and merely corresponds to the facilitated processing of the presented words due to previous presentation of the corresponding concepts. To exclude this possibility, we used paired t tests to compare the BOLD signal corresponding to new concrete words and the matching concrete words which the subjects failed to retrieve. In the majority of the clusters where specific deactivations during incidental retrieval were found, the BOLD amplitude for the failed items did not significantly differ from the new concrete items (p > .1), supporting the conclusion that the reductions were associated with incidental retrieval, not priming. Examining the effect sizes showed that although the clusters did show small to moderate-sized effects for failed compared to new items (d = 0.28 to d = 0.34), the 95% confidence intervals for the effect size in all of these cases encompassed zero. Thus, it is justified to assume that these clusters were primarily modulated by incidental retrieval, not the previous presentation of the corresponding concepts.

Only two clusters differed from this pattern: In parahippocampal cortex, the BOLD amplitude showed a trend toward a linear decrease (new concrete > failed > retrieved), with the failed items significantly different from the new concrete items [t(13) = 3.12, p < .01; d = 0.6; 95% CI = 0.2 to 1.4], whereas the comparison between the failed and the successfully retrieved items just failed to reach significance [t(13) = 1.99, p < .07; d = 1.2; 95% CI = −0.04 to 1.1). Another region which deviated from the overall pattern was the left superior temporal gyrus, which exhibited nongraded sensitivity to stimulus oldness: the failed items differed from new items [t(13) = 2.7, p = .02; d = 0.4; 95% CI = 0.12 to 1.3], whereas they did not significantly differ from the retrieved items [t(13) = 1.6, p > .1; d = 0.75; 95% CI = −0.14 to 0.96].

Additionally, we re-examined the behavioral data to test whether categorization of the words which matched the encoded items (both incidentally retrieved as well as those which the subjects failed to retrieve) occurred faster compared to categorization of the new concrete words in the incidental condition, which would suggest that the observed neural response merely served to facilitate semantic categorization and occurred independently of subsequent episodic retrieval. We found no behavioral evidence of priming: The semantic categorization response for the incidentally retrieved items was not faster compared to the new concrete items [retrieved: 1121 msec; new: 1194 msec; t(22) = 0.82, p > .1]; a similar result was found when comparing the semantic categorization speed for the failed items with the new concrete items [failed: 1202 msec, new: 1194 msec; t(22) = 0.62, p > .1].

Thus, the reduced posterior BOLD response, which was specific for incidental retrieval, differed from an implicit priming response in that it was not related to stimulus oldness per se, nor did it predict faster semantic categorization of the items. Instead, it was specifically expressed for the cues that were related to the study items and led to successful retrieval. The only exception to this pattern was the left superior temporal gyrus (BA 22), which was sensitive to stimulus oldness, not incidental retrieval. Parahippocampal cortex showed a graded pattern of deactivation, such that the incidentally retrieved items led to strongest deactivation, whereas the items which the subjects failed to retrieve led to intermediate deactivation compared to retrieved and new items.

EXPERIMENT 2

In a second experiment on a new group of participants, the same experimental protocol as in Experiment 1 was used to record ERPs. The fMRI data showed that a neural response specific to incidental retrieval occurred in left extrastriate cortex, which suggests that information of stimulus oldness leading to incidental retrieval may have been available relatively early, already during the visual processing of the item within the first 200 msec after stimulus presentation. However, fMRI does not allow exact mapping of the time course of the neural events, thus the possibility that the observed distinct brain activation pattern was a result, not a precursor of incidental retrieval, cannot be ruled out on the basis of fMRI data alone. ERPs have a temporal resolution on the millisecond scale and are therefore suitable for testing the hypothesis of early processing differences in relation to incidental retrieval. In light of the fMRI findings, we specifically tested whether a response in left posterior cortex characterized incidental retrieval, and mapped the time course of this effect. In addition, we tested whether the P600 ERP component, found at 500–800 msec poststimulus with a left parietal distribution, and associated with episodic retrieval (Wilding & Rugg, 1996), is influenced by retrieval intention. Based on previous findings (Curran, 1999), we hypothesized that the P600 component is sensitive to retrieval success, and this effect does not interact with retrieval intention.

Subjects

In the EEG data collection, the subjects were 35 healthy, young right-handed adults (mean age = 23.1 ± 1.3 years, 16 women) with normal or corrected-to-normal vision and normal hearing. After the preprocessing (see below), all subjects with fewer than 10 items remaining in any stimulus category were rejected, resulting in 23 subjects. The subjects gave written informed consent prior to participation. The study was approved by the regional ethics committee of Western Norway.

Acquisition and Preprocessing of the ERP Data

The electrophysiological recordings were conducted in an electromagnetically shielded, sound-attenuated test chamber (Rainford, Wigan, UK) in the Department of Biological and Medical Psychology, University of Bergen, Norway. EEGs were recorded continuously, with a sampling rate of 1000 Hz, from 61 Ag/AgCl electrodes mounted in an elastic cap (EasyCap; Falk Minow Services, Breitenbrunn, Germany) placed according to the International 10–20 recording system. Eye movements were monitored by an electrode placed under the left eye, with the reference electrode placed at the nasion and the ground electrode placed at the right temple. Impedance of all electrodes was kept under 10 kΩ. The data were recorded for off-line processing using BrainVision Recorder software (BrainProducts, Munich, Germany). The data were re-referenced to common average, filtered using a 0.5–45 Hz bandpass filter, down-sampled to 250 Hz, and divided into 4092-msec epochs around the stimulus onset. The epochs were baseline corrected relative to the mean of 700 msec prestimulus period. Artifact rejection was performed using EEGLAB software (Delorme & Makeig, 2004). Epochs containing artifacts were rejected using a probability distribution-based metric (threshold 3 standard deviations). Subsequently, temporal independent component analysis of each subject's data was performed to identify and reject artifacts; components accounting for the eye movements and blinks, as well as muscle activity, were rejected. Finally, all epochs were denoised using the wavelet transform (Quiroga & Garcia, 2003).

Statistical Analysis of the ERP Data

As the fMRI data (see below) suggested that the signal leading to incidental retrieval is expressed in left occipital cortex, we concentrated the analysis on left posterior inferior electrode sites (averaging over electrodes P5, P7, PO3, PO7, O1) within the early range of visual potentials (100–200 msec). The amplitude of the waveforms was averaged within the time windows between 100–150 and 150–200 msec. For analysis, a repeated measures ANOVA (using SPSS 17.0 software) with 2 × 2 design (Task: intentional/incidental × Item: retrieved/new concrete) was performed.

The left-distributed P600 effect was tested in the left parietal electrode P3 in the time window 500–800 msec. The averaged amplitude was submitted, as above, to 2 × 2 repeated measures ANOVA (Task: intentional/incidental × Item: retrieved/new concrete).

Results

Behavioral Results

The behavioral data largely replicated the patterns found in the fMRI experiment. t Tests comparing the results under intentional and incidental conditions showed that the subjects recalled more items in the intentional compared to the incidental condition [intentional: 55.1%, incidental: 46.7%; t(22) = 3.48, p < .005]. The subjects chose “remember” as the first response compared to categorization more often in the intentional than in the incidental condition [intentional: 41%, incidental: 8%; t(22) = 4.4, p < .005]. There was a weak tendency in the recall performance toward being faster in the intentional condition [intentional: 1322 msec, incidental: 1358 msec; t(22) = 1.21, p = .2]. Categorization of new concrete items was slower in the intentional condition compared to the incidental condition [intentional: 1207 msec, incidental: 1093 msec; t(22) = 4.27, p < .005], whereas categorization speed for abstract items did not differ between the conditions, suggesting that new concrete items were evaluated for their “oldness” only under intentional condition.

ERP Results

In the analysis of the ERP data, we concentrated on left-sided occipito-temporal electrodes in two time windows after item onset (100–150 and 150–200 msec), thereby targeting the visual evoked potentials most likely to arise from extrastriate cortex (Di Russo, Martinez, Sereno, Pitzalis, & Hillyard, 2002). In the average waveforms (Figure 4A), we observed the typical visual evoked potentials with a positivity from 80 to 130 msec (P1) and a subsequent negativity from 160 to 190 msec (N1). Windowed measurements of the peak amplitudes of the P1 and N1 components were subjected to a 2 × 2 repeated measures ANOVA (Task: intentional/incidental × Item: retrieved/new concrete). No significant Task × Item interaction was found during the first time window, corresponding to the P1 component (p > .3), but such an interaction was seen in the second time window, corresponding to the N1 component [F(1, 22) = 4.95, p = .04]. Post hoc comparisons confirmed that the interaction was driven by the N1 amplitude increase for the incidentally retrieved items compared to the new concrete items [t(22) = 2.35, p = .03; d = 0.2; 95% CI = 0.1 to 0.9], whereas the N1 amplitude did not differ between the retrieved and the new concrete items in the intentional condition (p > .2; d = 0.1; 95% CI = −0.2 to 0.6) (Figure 4A). Source localization studies of the visual N1 component suggest that its generators are located in extrastriate cortex, close to human V4 (Di Russo et al., 2002; Gonzalez, Clark, Fan, Luck, & Hillyard, 1994); a location which is in good agreement with our fMRI findings.

Figure 4. 

ERP indices of intentional and incidental retrieval. (A) Incidentally retrieved items in the ERP data selectively increased the amplitude of the negative peak (N1) at ∼170 msec after item onset at the left parieto-occipital recording site compared to new concrete items. Such effect was not found in the intentional condition, where the retrieved and new items did not differ in N1 amplitude. (B) Scatterplot showing the linear relationship between N1 latency during incidental retrieval and response time to incidentally retrieved items. No such correlation was seen during the other three conditions. (C) The late left parietal ERP component (P600) previously found to index recollection was increased in both intentional and incidental conditions for retrieved compared to new concrete items. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words.

Figure 4. 

ERP indices of intentional and incidental retrieval. (A) Incidentally retrieved items in the ERP data selectively increased the amplitude of the negative peak (N1) at ∼170 msec after item onset at the left parieto-occipital recording site compared to new concrete items. Such effect was not found in the intentional condition, where the retrieved and new items did not differ in N1 amplitude. (B) Scatterplot showing the linear relationship between N1 latency during incidental retrieval and response time to incidentally retrieved items. No such correlation was seen during the other three conditions. (C) The late left parietal ERP component (P600) previously found to index recollection was increased in both intentional and incidental conditions for retrieved compared to new concrete items. INT = intentional retrieval; INCI = incidental retrieval; R = retrieved words; C = new concrete words.

As the modulation of N1 has been associated with cross-modal priming (Jemel, Pisani, Rousselle, Crommelinck, & Bruyer, 2005), we performed a re-examination of the data, as was done in the fMRI experiment. The items which the subjects failed to retrieve in the incidental condition did not modulate the amplitude of N1 compared to the new concrete items (p > .5; d = 0.05; 95% CI = −0.3 to 0.5), a pattern which is similar to the characteristics of the response observed in extrastriate cortex in the fMRI study. This confirms that the N1 amplitude modulation was associated with incidental retrieval, not stimulus oldness. We also tested whether there was behavioral evidence for priming (i.e., reduced semantic categorization reaction time), which could be an alternative explanation for the observed ERP effect. Within the incidental condition, the categorization speed did not significantly differ between the retrieved and the new concrete items [retrieved: 1019 msec, new: 1093 msec; t(22) = 1.57, p > .1], nor between the failed and the new concrete items [failed: 1131 msec, new: 1093 msec; t(22) = 1.74, p > .1].

Finally, assuming that the modulation of N1 amplitude marks the initiation of processes resulting in incidental retrieval, it could be expected that the latency of the N1 peak predicts the response time of incidental retrieval, but no such relationship can be seen in intentional retrieval. A correlation analysis found a significant correlation between N1 latency and response time for the incidental (r = .40, p < .05), but not intentional retrieval (r = .20, p = .30; Figure 4B). Also, there was no significant correlation between N1 latency and the response time for the new concrete items in either condition (p > .1). This selective relationship between N1 latency and speed of incidental retrieval further supports the notion that processes leading to incidental retrieval began at the time window corresponding to the N1 component.

In further agreement with the results from the fMRI study, the ERP data also showed similarities between intentional and incidental retrieval. Several previous ERP studies of episodic retrieval have observed a left parietal old/new effect (Duzel, Yonelinas, Mangun, Heinze, & Tulving, 1997) which has been found to be unaffected by retrieval intention (Curran, 1999). We tested for this effect in the left parietal electrode site in the time window between 500 and 800 msec, using a 2 × 2 (Task × Item) repeated measures ANOVA. We found a significant main effect of item [retrieved > new concrete; F(1, 22) = 8.76, p = .007; Figure 4B]; no other effect was significant. This result agrees with previous findings in suggesting that the left parietal old/new effect expressed by the P600 component is not modulated by retrieval intention (Curran, 1999).

DISCUSSION

In this study, we explored the brain activity associated with episodic retrieval in the absence of retrieval intention. Our aim was to examine whether early processing of potential retrieval cues can provide bottom–up signals, which lead to episodic retrieval in the absence of prestimulus top–down control. By using fMRI (Experiment 1), we distinguished between brain activity which (1) contributed only to incidental retrieval (bottom–up input); (2) was commonly associated with successful retrieval in both intentional and incidental conditions (retrieval success); and (3) was expressed for all items in the intentional task, regardless of success (retrieval intention). Of these findings, the novel contribution of this study was isolating the brain activity contributing to incidental retrieval. We subsequently used ERPs (Experiment 2) to confirm that such selective brain activity was evident during the early stages of stimulus processing, occurring prior to indices of episodic retrieval and thus corresponding to bottom–up input to episodic memory system. These findings provide novel evidence that incidental or automatic (Osada et al., 2008) retrieval of episodic memories can be triggered by signals from posterior cortex, including the extrastriate and parahippocampal cortices, as early as within the first 200 msec after the presentation of visual word stimuli.

Several of the brain regions implicated here in providing the input to initiate incidental retrieval in response to visual cues (left-lateralized extrastriate cortex, parahippocampal cortex, and posterior cingulate cortex) have strong anatomical connections with each other as well as with the hippocampus (Lavenex & Amaral, 2000; Suzuki & Amaral, 1994; Vogt, Rosene, & Pandya, 1979), making them suitable candidates for eliciting the signals that evoke memories incidentally. Similarly to the graded activity pattern found here (retrieved < failed < new), parahippocampal cortex has been found to exhibit graded decline corresponding to the increase in the perceived memory strength (Gonsalves, Kahn, Curran, Norman, & Wagner, 2005); a graded pattern of parahippocampal cortex activation has also been found to signal increased accessibility of mnemonic information to conscious processing (Habib & Nyberg, 2008). In an interesting parallel to the present findings, parahippocampal cortex has been associated with reacting to contextually distinctive stimuli (Strange, Otten, Josephs, Rugg, & Dolan, 2002), as well as indexing representations which are in the focus of consciousness (Cabeza et al., 2003). Extrastriate cortex, which has been associated with amodal conceptual processing, has reciprocal connections with parahippocampal cortex. This suggests that incidental retrieval was initiated as a result of oldness detection on a conceptual level, performed in the interaction between extrastriate and parahippocampal cortices. Posterior cingulate cortex has been implicated in the retrieval from episodic memory and hypothesized to perform preliminary monitoring of the recollected information (Yonelinas, Otten, Shaw, & Rugg, 2005). Finally, an extensive deactivation was found in the anterior part of medial parietal cortex, corresponding to the paracentral lobule. The functional significance of this finding is not clear. Deactivation of the right paracentral lobule has been found to accompany increasing difficulty in a working memory task (Sweet et al., 2008), and is associated with the “default mode” network, which is deactivated during cognitively demanding tasks (Mazoyer et al., 2001). Thus, this deactivation may have reflected the additional processing requirements associated with items which were accompanied by incidental retrieval. However, the activation we found is close to the anterior medial parietal area, which has been found to deactivate during episodic retrieval (Buckner, Raichle, Miezin, & Petersen, 1996), suggesting that the deactivation of this area may have a particular significance for memory processes.

In agreement with the fMRI finding of selective response to incidental retrieval in extrastriate cortex, the ERP results (Experiment 2) showed a selective increase for incidentally retrieved items in the amplitude of N1, a negative-going visual potential originating from extrastriate cortex (Di Russo et al., 2002; Gonzalez et al., 1994). This confirms that the neural activity selectively predicting retrieval in the absence of top–down control corresponds to distinctions occurring relatively early (<200 msec) during stimulus processing. This finding is in agreement with a number of recent findings pointing to the possibility that processes contributing to explicit (i.e., conscious) retrieval begin within this time frame (Ecker, Zimmer, Groh-Bordin, & Mecklinger, 2007; Speer & Curran, 2007). In particular, activity in extrastriate areas within 150 msec after stimulus onset has been found to contribute to associative recognition memory (Duzel, Habib, Guderian, & Heinze, 2004).

The key signature of incidental retrieval in Experiment 1 consisted of relative BOLD signal reductions compared to the processing of new items. A reduced neural response to a repeated presentation of an item has been related to implicit memory, or priming (Henson, 2003). In a similar vein, the modulation of N1 ERP component, as observed in Experiment 2, has been associated with cross-modal priming (Jemel et al., 2005). This raises the possibility that the observed responses to visually presented words occurred independently of episodic retrieval and signaled merely the fact that subjects had accessed the corresponding concepts during encoding. We consider this possibility unlikely for several reasons. First, although implicit memory—often perceptual (Voss, Baym, & Paller, 2008; Habib & Nyberg, 1997), but also conceptual (Wolk et al., 2004)—has been demonstrated to influence the performance on explicit memory tasks, this seems to be limited to familiarity-based judgments which do not draw upon recollection of contextual details (Voss & Paller, 2009; Rajaram & Geraci, 2000). A plausible mechanism of such influence is an interaction between processing fluency and performance monitoring, without the involvement of the episodic memory system (Wolk et al., 2004). By contrast, in both of our experiments, the items which the subjects reported as incidentally retrieved were accompanied by similar neural signatures of episodic retrieval success as those items which were intentionally retrieved (see below). This argues against the possibility that the subjects' responses were based on evaluating the relative fluency of word processing, rather than the recollective experience. Second, we did not find any behavioral evidence of priming; the semantic categorization of the items which had been previously presented did not occur faster compared to the new items. Third, modulation of both BOLD amplitude as well as the N1 component amplitude predicting incidental retrieval was selectively expressed for the items which were retrieved, not for all the items which had been previously presented. In Experiment 1, the only region which was sensitive to stimulus oldness in the incidental condition was situated in the left superior temporal gyrus, which has been associated with cross-modal priming (Badgaiyan, Schacter, & Alpert, 2001), possibly supporting binding across different modalities. In the present experiment, this may have signaled increased fluency in the integration between the lexical representation of the visually presented word and the concept which it signified, due to the fact that the corresponding concepts had been activated during encoding. However, this response was not sufficient to evoke incidental retrieval; instead, a number of other regions expressed selective activation predicting retrieval. Our findings support the argument (Brown & Aggleton, 2001; Brown & Xiang, 1998) that although priming and familiarity discrimination in the service of declarative memory may share neural mechanisms (i.e., deactivation of neural activity in regions associated with stimulus processing), it does not render these phenomena functionally equivalent.

In contrast to the findings in the incidental condition, the retrieval success in the intentional condition was accompanied by a selective increase in the activation of predominantly right-lateralized parietal and perirhinal cortex. The dissociation within the medial temporal lobe depending on the retrieval intention suggests that the task set which the subjects have adopted influences already the basic processing of environmental stimuli, including the locus where their oldness is detected in such form so as to be useful for recollection (cf. Busch, Groh-Bordin, Zimmer, & Herrmann, 2008; Gruber & Muller, 2006; Henson, Shallice, Gorno-Tempini, & Dolan, 2002).

Despite the observed differences in brain responses during intentional and incidental retrieval, successful retrieval—regardless of intentionality—was associated with similar levels of posterior hippocampus activation, as well as similar amplitude increases of the P600 ERP component (Duzel et al., 1997). Additionally, successful retrieval was associated with the engagement of precuneus, ventrolateral prefrontal cortex (vlPFC), and ACC. Similar findings have been consistently associated with retrieval success. The hippocampus has been suggested to act as a “pointer” which binds together the components of the multimodal episodic memory trace (Eldridge et al., 2000), whereas the precuneus has been associated with representation of the retrieved information, or accumulation of evidence for a mnemonic decision (Wagner, Shannon, Kahn, & Buckner, 2005). Similarly, the P600 component has been suggested to index the interactions between medial temporal lobe and parietal cortex during episodic retrieval (Wilding & Rugg, 1996). The exact role of vlPFC and ACC in episodic retrieval is less clear. vlPFC and ACC activation is a common finding in retrieval tasks (Cabeza & Nyberg, 2000), notably even under experimental conditions fostering incidental retrieval during tasks which do not draw upon episodic memory (Smith, Henson, Dolan, & Rugg, 2004; Maratos, Dolan, Morris, Henson, & Rugg, 2001; Duzel et al., 1999). A possible explanation for the role of these brain regions is postretrieval processing—performing operations which are contingent upon the retrieved information.

We note that caution is warranted when interpreting the findings associated to retrieval success. As the items which the subjects reported as recollected were associated with two button presses, whereas the new items were associated with a single button press, our analysis could not separate the results associated with retrieval success and the results associated with non-mnemonic factors, such as the increase in the number of decisions per item. Given the extensive evidence linking the hippocampus, the precuneus, and the P600 ERP component to recollection, we do not believe that these findings in our data resulted from non-mnemonic factors. However, we cannot exclude the possibility that the increased demands in decision-making or organizing the response sequence may have influenced the engagement of vlPFC and ACC.

Finally, in keeping with many prior studies (Lepage et al., 2000; Tomita et al., 1999), both successfully retrieved as well as new concrete items in the intentional condition were associated with increased activation of right dorsolateral and left anterior prefrontal cortex. Thus, this activation was related to intention to retrieve, but independent of retrieval success, a finding which agrees with the view of these brain regions' importance for strategic, top–down retrieval control (Rugg et al., 1997). In addition to that, we found increased activation in parietal cortex and cerebellum, regions both associated with participating in cognitive control (Bellebaum & Daum, 2007; Egner & Hirsch, 2005; Culham & Kanwisher, 2001).

Conclusions

Taken together, the results from this study suggest that accessing episodic memories can be initiated both via top–down signals from prefrontal cortex as well as via bottom–up signals in a “low route,” arising during the early processing of a stimulus. Once initiated, the subsequent processing follows similar paths in both cases, engaging the hippocampus to retrieve the corresponding episodic information. Thus, intentionally and incidentally retrieved memories overlap with regard to the brain activity which subserves actual retrieval success as well as postretrieval processing, but diverge in how they are initiated. This difference is qualitative: An early indicator of sufficient strength that a stimulus has a matching trace in the long-term memory is necessary for the incidental retrieval, whereas the top–down guided assessment of all cues for a possible match is performed for the intentional retrieval. Our findings are in agreement with the general framework proposed by Moscovitch (1992), according to which the hippocampus, the neural “module” supporting episodic retrieval, provides automatic output when information resulting from cue processing exceeds a certain threshold, with prefrontal processes providing support and organization. We conclude that the early neural signals described in the present study represent a “low route” via which stimuli can evoke episodic memories without retrieval intention and prefrontal top–down modulation.

Acknowledgments

This study was supported by a grant from NOS-HS for the Nordic Center of Excellence in Cognitive Control. We thank Edvard Moser, Rodrigo Quian Quiroga, and Daniel L. Schacter for comments on earlier versions of the manuscript, as well as Karolina Kauppi and Rune Eikeland for assistance in data collection.

Reprint requests should be sent to Kristiina Kompus, Department of Biological and Medical Psychology, University of Bergen, Jonas Lies vei 91, N-5009 Bergen, Norway, or via e-mail: kristiina.kompus@psybp.uib.no.

Note

1. 

Our distinction between “incidental” and “intentional” episodic retrieval refers to variations in strategic processes which support episodic retrieval; in both cases, the resulting episodic memory is experienced consciously. Thus, this distinction is not to be confused with the occasional usage of the term “incidental memory” to collectively denote various implicit memory processes, such as priming, or testing conditions to study implicit memory processes (see, e.g., Richardson-Klavehn, Gardiner, & Java, 1996).

REFERENCES

REFERENCES
Badgaiyan
,
R. D.
,
Schacter
,
D. L.
, &
Alpert
,
N. M.
(
1999
).
Auditory priming within and across modalities: Evidence from positron emission tomography.
Journal of Cognitive Neuroscience
,
11
,
337
348
.
Badgaiyan
,
R. D.
,
Schacter
,
D. L.
, &
Alpert
,
N. M.
(
2001
).
Priming within and across modalities: Exploring the nature of rCBF increases and decreases.
Neuroimage
,
13
,
272
282
.
Bellebaum
,
C.
, &
Daum
,
I.
(
2007
).
Cerebellar involvement in executive control.
Cerebellum
,
6
,
184
192
.
Berntsen
,
D.
(
1996
).
Involuntary autobiographical memories.
Applied Cognitive Psychology
,
10
,
435
454
.
Brown
,
M. W.
, &
Aggleton
,
J. P.
(
2001
).
Recognition memory: What are the roles of the perirhinal cortex and hippocampus?
Nature Reviews Neuroscience
,
2
,
51
61
.
Brown
,
M. W.
, &
Xiang
,
J. Z.
(
1998
).
Recognition memory: Neuronal substrates of the judgement of prior occurrence.
Progress in Neurobiology
,
55
,
149
189
.
Buckner
,
R. L.
,
Raichle
,
M. E.
,
Miezin
,
F. M.
, &
Petersen
,
S. E.
(
1996
).
Functional anatomic studies of memory retrieval for auditory words and visual pictures.
Journal of Neuroscience
,
16
,
6219
6235
.
Busch
,
N. A.
,
Groh-Bordin
,
C.
,
Zimmer
,
H. D.
, &
Herrmann
,
C. S.
(
2008
).
Modes of memory: Early electrophysiological markers of repetition suppression and recognition enhancement predict behavioral performance.
Psychophysiology
,
45
,
25
35
.
Cabeza
,
R.
,
Dolcos
,
F.
,
Prince
,
S. E.
,
Rice
,
H. J.
,
Weissman
,
D. H.
, &
Nyberg
,
L.
(
2003
).
Attention-related activity during episodic memory retrieval: A cross-function fMRI study.
Neuropsychologia
,
41
,
390
399
.
Cabeza
,
R.
, &
Nyberg
,
L.
(
2000
).
Imaging cognition: II. An empirical review of 275 PET and fMRI studies.
Journal of Cognitive Neuroscience
,
12
,
1
47
.
Cohen
,
N. J.
, &
Squire
,
L. R.
(
1981
).
Retrograde-amnesia and remote memory impairment.
Neuropsychologia
,
19
,
337
356
.
Conway
,
M. A.
, &
Pleydell-Pearce
,
C. W.
(
2000
).
The construction of autobiographical memories in the self-memory system.
Psychological Review
,
107
,
261
288
.
Culham
,
J. C.
, &
Kanwisher
,
N. G.
(
2001
).
Neuroimaging of cognitive functions in human parietal cortex.
Current Opinion in Neurobiology
,
11
,
157
163
.
Curran
,
T.
(
1999
).
The electrophysiology of incidental and intentional retrieval: ERP old new effects in lexical decision and recognition memory.
Neuropsychologia
,
37
,
771
785
.
Delorme
,
A.
, &
Makeig
,
S.
(
2004
).
EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis.
Journal of Neuroscience Methods
,
134
,
9
21
.
Di Russo
,
F.
,
Martinez
,
A.
,
Sereno
,
M. I.
,
Pitzalis
,
S.
, &
Hillyard
,
S. A.
(
2002
).
Cortical sources of the early components of the visual evoked potential.
Human Brain Mapping
,
15
,
95
111
.
Duzel
,
E.
,
Cabeza
,
R.
,
Picton
,
T. W.
,
Yonelinas
,
A. P.
,
Scheich
,
H.
,
Heinze
,
H. J.
,
et al
(
1999
).
Task-related and item-related brain processes of memory retrieval.
Proceedings of the National Academy of Sciences, U.S.A.
,
96
,
1794
1799
.
Duzel
,
E.
,
Habib
,
R.
,
Guderian
,
S.
, &
Heinze
,
H. J.
(
2004
).
Four types of novelty–familiarity responses in associative recognition memory of humans.
European Journal of Neuroscience
,
19
,
1408
1416
.
Duzel
,
E.
,
Yonelinas
,
A. P.
,
Mangun
,
G. R.
,
Heinze
,
H. J.
, &
Tulving
,
E.
(
1997
).
Event-related brain potential correlates of two states of conscious awareness in memory.
Proceedings of the National Academy of Sciences, U.S.A.
,
94
,
5973
5978
.
Ecker
,
U. K. H.
,
Zimmer
,
H. D.
,
Groh-Bordin
,
C.
, &
Mecklinger
,
A.
(
2007
).
Context effects on familiarity are familiarity effects of context: An electrophysiological study.
International Journal of Psychophysiology
,
64
,
146
156
.
Egner
,
T.
, &
Hirsch
,
J.
(
2005
).
The neural correlates and functional integration of cognitive control in a Stroop task.
Neuroimage
,
24
,
539
547
.
Eldridge
,
L. L.
,
Knowlton
,
B. T.
,
Furmanski
,
C. S.
,
Bookheimer
,
S. Y.
, &
Engel
,
S. A.
(
2000
).
Remembering episodes: A selective role for the hippocampus during retrieval.
Nature Neuroscience
,
3
,
1149
1152
.
Fletcher
,
P. C.
, &
Henson
,
R. N. A.
(
2001
).
Frontal lobes and human memory: Insights from functional neuroimaging.
Brain
,
124
,
849
881
.
Gonsalves
,
B. D.
,
Kahn
,
I.
,
Curran
,
T.
,
Norman
,
K. A.
, &
Wagner
,
A. D.
(
2005
).
Memory strength and repetition suppression: Multimodal imaging of medial temporal cortical contributions to recognition.
Neuron
,
47
,
751
761
.
Gonzalez
,
C. M. G.
,
Clark
,
V. P.
,
Fan
,
S.
,
Luck
,
S. J.
, &
Hillyard
,
S. A.
(
1994
).
Sources of attention-sensitive visual event-related potentials.
Brain Topography
,
7
,
41
51
.
Gruber
,
T.
, &
Muller
,
M. M.
(
2006
).
Oscillatory brain activity in the human EEG during indirect and direct memory tasks.
Brain Research
,
1097
,
194
204
.
Habib
,
R.
, &
Nyberg
,
L.
(
1997
).
Incidental retrieval processes influence explicit test performance with data-limited cues.
Psychonomic Bulletin & Review
,
4
,
130
133
.
Habib
,
R.
, &
Nyberg
,
L.
(
2008
).
Neural correlates of availability and accessibility in memory.
Cerebral Cortex
,
18
,
1720
1726
.
Hall
,
N. M.
,
Gjedde
,
A.
, &
Kupers
,
R.
(
2008
).
Neural mechanisms of voluntary and involuntary recall: A PET study.
Behavioural Brain Research
,
186
,
261
272
.
Henson
,
R. N. A.
(
2003
).
Neuroimaging studies of priming.
Progress in Neurobiology
,
70
,
53
81
.
Henson
,
R. N. A.
,
Shallice
,
T.
,
Gorno-Tempini
,
M. L.
, &
Dolan
,
R. J.
(
2002
).
Face repetition effects in implicit and explicit memory tests as measured by fMRI.
Cerebral Cortex
,
12
,
178
186
.
Jemel
,
B.
,
Pisani
,
M.
,
Rousselle
,
L.
,
Crommelinck
,
M.
, &
Bruyer
,
R.
(
2005
).
Exploring the functional architecture of person recognition system with event-related potentials in a within- and cross-domain self-priming of faces.
Neuropsychologia
,
43
,
2024
2040
.
Lavenex
,
P.
, &
Amaral
,
D. G.
(
2000
).
Hippocampal–neocortical interaction: A hierarchy of associativity.
Hippocampus
,
10
,
420
430
.
Lepage
,
M.
,
Ghaffar
,
O.
,
Nyberg
,
L.
, &
Tulving
,
E.
(
2000
).
Prefrontal cortex and episodic memory retrieval mode.
Proceedings of the National Academy of Sciences, U.S.A.
,
97
,
506
511
.
Maratos
,
E. J.
,
Dolan
,
R. J.
,
Morris
,
J. S.
,
Henson
,
R. N. A.
, &
Rugg
,
M. D.
(
2001
).
Neural activity associated with episodic memory for emotional context.
Neuropsychologia
,
39
,
910
920
.
Mazoyer
,
B.
,
Zago
,
L.
,
Mellet
,
E.
,
Bricogne
,
S.
,
Etard
,
O.
,
Houde
,
O.
,
et al
(
2001
).
Cortical networks for working memory and executive functions sustain the conscious resting state in man.
Brain Research Bulletin
,
54
,
287
298
.
Milner
,
B.
,
Corkin
,
S.
, &
Teuber
,
H. L.
(
1968
).
Further analysis of hippocampal amnesic syndrome: 14-Year follow-up study of Hm.
Neuropsychologia
,
6
,
215
234
.
Miyashita
,
Y.
(
2004
).
Cognitive memory: Cellular and network machineries and their top–down control.
Science
,
306
,
435
440
.
Moscovitch
,
M.
(
1992
).
Memory and working-with-memory: A component process model based on modules and central systems.
Journal of Cognitive Neuroscience
,
4
,
257
267
.
Naya
,
Y.
,
Yoshida
,
M.
, &
Miyashita
,
Y.
(
2001
).
Backward spreading of memory-retrieval signal in the primate temporal cortex.
Science
,
291
,
661
664
.
Nyberg
,
L.
,
McIntosh
,
A. R.
,
Houle
,
S.
,
Nilsson
,
L. G.
, &
Tulving
,
E.
(
1996
).
Activation of medial temporal structures during episodic memory retrieval.
Nature
,
380
,
715
717
.
Nyberg
,
L.
,
Tulving
,
E.
,
Habib
,
R.
,
Nilsson
,
L. G.
,
Kapur
,
S.
,
Houle
,
S.
,
et al
(
1995
).
Functional brain maps of retrieval mode and recovery of episodic information.
NeuroReport
,
7
,
249
252
.
Osada
,
T.
,
Adachi
,
Y.
,
Kimura
,
H. M.
, &
Miyashita
,
Y.
(
2008
).
Towards understanding of the cortical network underlying associative memory.
Philosophical Transactions of the Royal Society of London, Series B, Biological Sciences
,
363
,
2187
2199
.
Quiroga
,
R. Q.
, &
Garcia
,
H.
(
2003
).
Single-trial event-related potentials with wavelet denoising.
Clinical Neurophysiology
,
114
,
376
390
.
Rajaram
,
S.
, &
Geraci
,
L.
(
2000
).
Conceptual fluency selectively influences knowing.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
26
,
1070
1074
.
Richardson-Klavehn
,
A.
, &
Gardiner
,
J. M.
(
1995
).
Retrieval volition and memorial awareness in stem completion: An empirical analysis.
Psychological Research
,
57
,
166
178
.
Richardson-Klavehn
,
A.
,
Gardiner
,
J. M.
, &
Java
,
R. I.
(
1996
).
Memory: Task dissociations, process dissociations, and dissociations of consciousness.
In G. Underwood (Ed.),
Implicit cognition
(pp.
85
158
).
New York
:
Oxford University Press
.
Rugg
,
M. D.
,
Fletcher
,
P. C.
,
Frith
,
C. D.
,
Frackowiak
,
R. S. J.
, &
Dolan
,
R. J.
(
1997
).
Brain regions supporting intentional and incidental memory: A PET study.
NeuroReport
,
8
,
1283
1287
.
Rugg
,
M. D.
,
Mark
,
R. E.
,
Walla
,
P.
,
Schloerscheidt
,
A. M.
,
Birch
,
C. S.
, &
Allan
,
K.
(
1998
).
Dissociation of the neural correlates of implicit and explicit memory.
Nature
,
392
,
595
598
.
Schacter
,
D. L.
,
Badgaiyan
,
R. D.
, &
Alpert
,
N. M.
(
1999
).
Visual word stem completion priming within and across modalities: A PET study.
NeuroReport
,
10
,
2061
2065
.
Schacter
,
D. L.
, &
Wagner
,
A. D.
(
1999
).
Medial temporal lobe activations in fMRI and PET studies of episodic encoding and retrieval.
Hippocampus
,
9
,
7
24
.
Smith
,
A. P. R.
,
Henson
,
R. N. A.
,
Dolan
,
R. J.
, &
Rugg
,
M. D.
(
2004
).
fMRI correlates of the episodic retrieval of emotional contexts.
Neuroimage
,
22
,
868
878
.
Speer
,
N. K.
, &
Curran
,
T.
(
2007
).
ERP correlates of familiarity and recollection processes in visual associative recognition.
Brain Research
,
1174
,
97
109
.
Strange
,
B. A.
,
Otten
,
L. J.
,
Josephs
,
O.
,
Rugg
,
M. D.
, &
Dolan
,
R. J.
(
2002
).
Dissociable human perirhinal, hippocampal, and parahippocampal roles during verbal encoding.
Journal of Neuroscience
,
22
,
523
528
.
Suzuki
,
W. A.
, &
Amaral
,
D. G.
(
1994
).
Topographic organization of the reciprocal connections between the monkey entorhinal cortex and the perirhinal and parahippocampal cortices.
Journal of Neuroscience
,
14
,
1856
1877
.
Sweet
,
L. H.
,
Paskavitz
,
J. F.
,
Haley
,
A. P.
,
Gunstad
,
J. J.
,
Mulligan
,
R. C.
,
Nyalakanti
,
P. K.
,
et al
(
2008
).
Imaging phonological similarity effects on verbal working memory.
Neuropsychologia
,
46
,
1114
1123
.
Tomita
,
H.
,
Ohbayashi
,
M.
,
Nakahara
,
K.
,
Hasegawa
,
I.
, &
Miyashita
,
Y.
(
1999
).
Top–down signal from prefrontal cortex in executive control of memory retrieval.
Nature
,
401
,
699
703
.
Tulving
,
E.
(
1983
).
Ecphoric processes in episodic memory.
Philosophical Transactions of the Royal Society of London, Series B, Biological Sciences
,
302
,
361
371
.
Vogt
,
B. A.
,
Rosene
,
D. L.
, &
Pandya
,
D. N.
(
1979
).
Thalamic and cortical afferents differentiate anterior from posterior cingulate cortex in the monkey.
Science
,
204
,
205
207
.
Voss
,
J. L.
,
Baym
,
C. L.
, &
Paller
,
K. A.
(
2008
).
Accurate forced-choice recognition without awareness of memory retrieval.
Learning and Memory
,
15
,
454
459
.
Voss
,
J. L.
, &
Paller
,
K. A.
(
2009
).
An electrophysiological signature of unconscious recognition memory.
Nature Neuroscience
,
12
,
349
355
.
Wagner
,
A. D.
,
Shannon
,
B. J.
,
Kahn
,
I.
, &
Buckner
,
R. L.
(
2005
).
Parietal lobe contributions to episodic memory retrieval.
Trends in Cognitive Sciences
,
9
,
445
453
.
Wheeler
,
M. A.
,
Stuss
,
D. T.
, &
Tulving
,
E.
(
1997
).
Toward a theory of episodic memory: The frontal lobes and autonoetic consciousness.
Psychological Bulletin
,
121
,
331
354
.
Wiggs
,
C. L.
, &
Martin
,
A.
(
1998
).
Properties and mechanisms of perceptual priming.
Current Opinion in Neurobiology
,
8
,
227
233
.
Wilding
,
E. L.
, &
Rugg
,
M. D.
(
1996
).
An event-related potential study of recognition memory with and without retrieval of source.
Brain
,
119
,
889
905
.
Wolk
,
D. A.
,
Schacter
,
D. L.
,
Berman
,
A. R.
,
Holcomb
,
P. J.
,
Daffner
,
K. R.
, &
Budson
,
A. E.
(
2004
).
An electrophysiological investigation of the relationship between conceptual fluency and familiarity.
Neuroscience Letters
,
369
,
150
155
.
Yonelinas
,
A. P.
,
Otten
,
L. J.
,
Shaw
,
K. N.
, &
Rugg
,
M. D.
(
2005
).
Separating the brain regions involved in recollection and familiarity in recognition memory.
Journal of Neuroscience
,
25
,
3002
3008
.