Process Dissociation between Contextual Retrieval and ... - CiteSeerX

10/28/2004

Process

Dissociation

between

Contextual

Retrieval

and

Item

Recognition Susanne Weisa, Karsten Spechtb, Peter Klavera, Indira Tendolkarc,d, Klaus Willmese, Jürgen Ruhlmannf, Christian E. Elgera, Guillén Fernándeza,d,g

a

Department of Epileptology, University of Bonn, 53105 Bonn, Germany; bInstitute of Medicine,

Research Centre Jülich, 52425 Jülich, Germany; cDepartment of Psychiatry, University Medical Center Nijmegen, 6500-HB Nijmegen, The Netherlands; dF.C. Donders Center for Cognitive Neuroimaging, University of Nijmegen, 6500-HB Nijmegen, The Netherlands; eSection of Neuropsychology, Department of Neurology, University Hospital Aachen, 52057 Aachen, Germany; fDepartment of Diagnostic and Therapeutic Neuroradiology, Medical Center Bonn, 53119 Bonn, Germany; gDepartment of Neurology, University Medical Center Nijmegen, University of Nijmegen, 6500-HB Nijmegen, The Netherlands

Short title: Contextual Retrieval and Item Recognition total number of characters: 19972

Correspondence:

Dr. Guillén Fernández F.C. Donders Center (181) P.O. Box 9101 6500 HB Nijmegen The Netherlands Phone Fax Email

+31-24-3610-749 +31-24-3610-652 [email protected]

Contextual Retrieval and Item Recognition

Page 2 of 14

Abstract We employed a source memory task in an event related functional MRI study to dissociate MTL processes associated with either contextual retrieval or item recognition. To introduce context during study, stimuli (photographs of buildings and natural landscapes) were transformed into one of four single-color-scales: red, blue, yellow, or green. In the subsequent old/new recognition memory test, all stimuli were presented as gray scale photographs, and oldresponses were followed by a four-alternative source judgment referring to the color in which the stimulus was presented during study. Our results suggest a clear-cut process dissociation within the human MTL. While an activity increase accompanies successful retrieval of contextual information, an activity decrease provides a familiarity signal that is sufficient for successful item recognition.

Key Words: declarative memory, familiarity, recollection, hippocampus, fMRI, source memory


Page 3 of 14

Introduction Dual-process models of recognition memory propose qualitative distinct forms of memory supporting recognition of an item [1,2]. While recollection, i.e. recognition of an item that is accompanied by contextual information, is truly episodic memory, recognition unaccompanied by contextual information can rely upon a sense of familiarity [1,3]. Lesion studies identified the medial temporal lobe (MTL) to be crucial for recognition memory [4], but the specific role of MTL subregions in contextual retrieval and item recognition is highly disputed [5,6]. Several event-related fMRI studies of memory formation have examined the difference in activity during encoding in different MTL subregions that leads to subsequent familiarity-based recognition as opposed to subsequent recollection [7,8]. These studies showed that encoding activity in the rhinal cortex selectively predicted familiarity-based recognition, whereas encoding activity in the hippocampus and posterior parahippocampal cortex selectively predicted recollection. Event-related fMRI studies of recognition memory investigating contextual retrieval showed that hippocampal activity increased with recollection success [9,10] (but see [11]). However, several studies investigating item recognition using of the old/new effect, the difference in brain activity between correctly recognized old, previously studied items (hits), and correctly identified new, previously unstudied items (correct rejections), did not reveal any MTL activity increases [11,12] (but see [6]). Electrophysiological studies have shown that anterior parahippocampal activity decreases during item recognition [13]. Also a recent meta-analysis of four event-related fMRI studies suggested that less anterior MTL activity is related to the amount of familiarity [14]. Thus, different functional processes, being either based on an activity increase or an activity decrease, might be involved with contextual retrieval and item memory in neighboring or overlapping MTL subregions.


Page 4 of 14

To study an activity decrease associated with successful item recognition, a reversed old/new contrast is not suited, since simple repetition effects would confound it. Contrary, a negative recognition effect (hits < misses (old, previously studied, items misclassified as new)) is less contaminated by repetition effects, because both items have been studied. This contrast revealed an anterior MTL effect, indicating that less activity in this brain region is related to item recognition success [12]. However, in that study no formal dissociation between recollection and item recognition was implemented. Therefore, in the present study, we investigated whether an MTL activity decrease is associated with simple, a-contextual item recognition and an MTL activity increase with associative, contextual retrieval.

Material and Methods Subjects: Twelve healthy volunteers (6 male; mean age: 28 y, range: 20-34 y) with normal or

corrected-to-normal vision participated in the experiment. Written informed consent was obtained in a manner approved by the Medical Ethics Committee of the University of Bonn and according to the Declaration of Helsinki (1991). Subjects were paid for their participation. Stimuli and Task: Stimuli consisted of 360 gray-scale photographs of either buildings or natural

landscapes (180 for each category) that were selected to be similar in complexity, brightness, and contrast. The experiment was divided into four study–test cycles. Between the study and test phases, there were short breaks of few minutes. During each of the four study phases, subjects saw 60 pictures which were shown transformed into red-, blue-, yellow- and green-scale (15 pictures for each color). Subjects were required to memorize each picture together with its color and to make a building-landscape decision. During each of the four recognition phases 90 pictures – 60 previously studied and 30 new - were presented as plain gray-scale photographs. Subjects were


Page 5 of 14

required to make an old / new decision for each picture and further to indicate the color in which the picture had been presented during study. To counter-balance stimuli across subjects, all pictures where randomly divided into three sets of 60 buildings and 60 landscapes each. For the study phases, two of the three sets of pictures were selected for each subject, resulting in four subjects seeing the same 240 pictures during the study phases. For each of these four subjects, different subsets of 60 pictures each were transformed into red-, blue-, yellow- and green-scale, so that no two subjects saw the pictures in the same color. During recognition, all subjects saw all stimuli in greay-scale. During the study phases, stimuli were presented sequentially for 800 ms with a randomized interstimulus interval (ISI) of 3600 to 5600 ms (mean 4600 ms). Additionally, 120 phases of baseline stimulation (i.e. black screen), each lasting for 2000 ms, were randomly intermixed as so-called “null events”. Both the way of ISI variation and the inclusion of null events have been shown to increase the statistical efficiency of event-related designs [15]. Subjects made the building-landscape decision by one of two alternative key-presses using the right hand. During the recognition phases, stimuli were shown randomly intermixed with 180 null events at the same presentation rate as during study. Subjects were required to make the old-new decision by one of two alternative key-presses using the right hand. For those stimuli judged as old, four colored squares were subsequently displayed on the screen for 800 ms. The interval between the offset of stimulus presentation and the onset of the color display presentation was varied between 1200 and 2400 ms (mean 1600 ms). Subjects were required to indicate the color in which the item had been presented during the study phase by one of four alternative key-presses using the right hand. fMRI Data Acquisition: Scans were performed on a 1.5 T scanner (Symphony, Siemens,

Erlangen, Germany) using standard gradients and a circular polarized phase array head coil. We acquired T2*-weighted axial EPI-scans parallel to the AC/PC line with the following parameters: number of slices (NS): 30; slice thickness (ST): 4 mm; interslice gap (IG): 0.4 mm;


Page 6 of 14

matrix size (MS): 64x64; field of view (FOV): 220 mm; echo time (TE): 50 ms; repetition time (TR): 2.95 s. T1-weighted 3D-FLASH scans were acquired for anatomical localization (NS: 120; ST: 1.5 mm; IG: none; MS: 256x256; FOV: 230 mm; TE: 4 ms; TR: 11 ms). fMRI Data Analysis: Statistical Parametric Mapping (SPM2, www.fil.ion.ucl.ac.uk) was used for

data analysis. Preprocessing using standard procedures included realignment, unwarping, slicetime correction, normalization into the stereotaxic Montreal Neurological Institute space, and spatial smoothing with an 8-mm FWHM isotropic Gaussian kernel. The time series data were band-pass filtered to remove artifacts occurring over time. The expected hemodynamic response at stimulus onset for each event-type was modeled by a canonical hemodynamic response function (HRF) and its temporal derivative. The temporal derivative was included to account for residual variance. The functions were convolved with the event-train of stimulus onsets to create covariates in a general linear model. Subsequently, parameter estimates of the HRF regressor for each of the different conditions were calculated from the least mean squares fit of the model to the time series. A random-effects group analysis was performed by entering parameter estimates for all conditions into a within-subject one-way ANOVA.

Results Behavioral Results: During study, the building/landscape decision task was performed with a mean accuracy of 94% (range: 90% - 98%). All trials with missing or incorrect building/landscape decisions were excluded from further analyses. Recognition memory performance and reaction times for the old-new decision are listed in Table 1. Accuracy of item recognition was assessed by the difference in probabilities of a correct old judgment and an old judgment for a new item (Pr = probability hit – probability false alarm). While recognition performance did not differ between stimulus classes (mean Prbuilding = 0.41 (SD: 0.14) versus mean Prlandscape = 0.40 (SD: 0.17), t11= 1.07, n.s.), it was well


Page 7 of 14

above the chance level of Pr = 0 (mean Pr = 0.41 (SD: 0.15), t11 = 8.40; p< 0.00001). The accuracy of source judgments was also well above the chance level of 25 % (mean correct = 50.8%, SD: 11%, t11 = 7.98, p