The Effect of Divided Attention on Encoding and Retrieval in Episodic Memory Revealed by Positron Emission Tomography Tetsuya Iidaka University of Toronto Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/12/2/267/1758705/089892900562093.pdf by guest on 18 May 2021 Nicole D. Anderson Princess Margaret Hospital Shitij Kapur Fukui Medical University Roberto Cabeza University of Alberta Fergus I. M. Craik University of Toronto Abstract & The effects of divided attention (DA) on episodic memory dorsolateral-prefrontal areas. Regions more active during encoding and retrieval were investigated in 12 normal young encoding than during retrieval were located in the hippocam- subjects by positron emission tomography (PET). Cerebral pus, temporal and the prefrontal cortex of the left hemisphere, blood flow was measured while subjects were concurrently and regions more active during retrieval than during encoding performing a memory task (encoding and retrieval of visually included areas in the medial and right-prefrontal cortex, basal presented word pairs) and an auditory tone-discrimination ganglia, thalamus, and cuneus. DA at encoding was associated task. The PET data were analyzed using multivariate Partial with specific decreases in rCBF in the left-prefrontal areas, Least Squares (PLS), and the results revealed three sets of whereas DA at retrieval was associated with decreased rCBF in neural correlates related to specific task contrasts. Brain a relatively small region in the right-prefrontal cortex. These activity, relatively greater under conditions of full attention different patterns of activity are related to the behavioral (FA) than DA, was identified in the occipital-temporal, medial, results, which showed a substantial decrease in memory and ventral-frontal areas, whereas areas showing relatively performance when the DA task was performed at encoding, more activity under DA than FA were found in the cerebellum, but no change in memory levels when the DA task was temporo-parietal, left anterior-cingulate gyrus, and bilateral performed at retrieval. & INTRODUCTION two processes because differences in performance Recent advances in functional neuroimaging techniques could be due to differences in encoding or retrieval. enable us to measure activity in the human brain while Neuroimaging studies provide the unique opportunity the subject is listening, thinking, paying attention, or to observe encoding and retrieval processes separately. remembering. In particular, memory functions in normal Previous studies of episodic memory using verbal mate- subjects have been investigated intensively by using rials indicate that left-prefrontal regions of the brain are positron emission tomography (PET), and these studies predominantly activated during episodic memory enco- have shed light on the neural correlates of encoding and ding; and, conversely, right-prefrontal regions are pre- retrieval processes in episodic memory (for reviews, see dominantly activated during episodic retrieval (Fletcher Cabeza & Nyberg, 1997; Fletcher, Frith, & Rugg, 1997). et al., 1995; Shallice et al., 1994; Tulving, Kapur, Craik, In behavioral studies of recall and recognition, it is Moscovitch, & Houle, 1994). In contrast, Kelley et al. impossible to disentangle the separate effects of the (1998) and Wagner et al. (1998) reported that the right- D 2000 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 12:2, pp. 267±280 Downloaded from http://www.mitpressjournals.org/doi/pdfplus/10.1162/089892900562093 by guest on 25 September 2021 prefrontal area is activated during encoding of nonverbal In the present study, subjects performed a memory materials. task and a secondary task concurrently; the memory task One of the main purposes of the present study was to involved encoding or retrieval of visually presented word examine the effects of divided attention (DA) on enco- pairs and verbal responses, whereas the secondary task ding and retrieval processes in human episodic memory involved easy or difficult auditory tone discrimination by using PET. When subjects are engaged in two differ- and manual responses. Based on previous results for ent tasks at once, they have to divide attention between verbal-memory tasks, we expected to see left-prefrontal the tasks and allocate mental resources to each task. activation associated with FA encoding (Tulving et al., Behavioral studies have shown that DA at encoding is 1994), and that this activation would be reduced in the associated with a substantial decrement in later memory DA condition (Fletcher et al., 1995; Shallice et al., 1994). Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/12/2/267/1758705/089892900562093.pdf by guest on 18 May 2021 performance, whereas DA at retrieval results in compara- Similarly, we expected to find right-prefrontal activation tively slight declines in memory (Anderson, Craik, & associated with FA retrieval, and given that DA at Naveh-Benjamin, 1998; Craik, Govoni, Naveh-Benjamin, retrieval has only small effects on memory performance, & Anderson, 1996; Baddeley, Lewis, Eldridge, & Thom- we speculated that deactivation due to DA might be son, 1984). This pattern of results was found in cases small. where secondary task costs were equivalent for enco- ding and retrieval, so it is not the case that retrieval is RESULTS simply protected at the expense of the concurrent task. It seems, rather, that whereas encoding is easily dis- Behavioral Data rupted by other ongoing processes, retrieval is more or Table 1 shows cued recall performance and mean less obligatory once the retrieval cue is provided or reaction times in the tone-discrimination task in the generated (Craik et al., 1996). An important question four experimental conditions. Recall performance was in this regard is whether there exists one common pool lowered substantially (0.79 to 0.58) when attention of attentional resources that can be allocated differen- was divided at encoding, but recall was affected only tially to various mental operations, depending on the minimally (0.78 to 0.75) when attention was divided relative importance of each operation. Some studies at retrieval. A two-way analysis of variance conducted have suggested an affirmative answer to this question. on cued recall performance showed a main effect of For example, event-related potential studies using the encoding/retrieval, F(1,11)=7.31, p<.05, a main effect dual-task paradigm have shown that there exists a of attention (FA vs. DA), F(1,11)=13.46, p<.01, and a reciprocity in the availability of resources for the primary significant interaction between attention and enco- task (visual detection) and the secondary task (auditory ding/retrieval, F(1,11)=11.15, p<.01. This is similar detection), and the allocation of resources may vary to the pattern reported by Craik et al. (1996). The according to the relative difficulty of the tasks (Wickens, tone-task data show that response times were slower Kramer, Vanasse, & Donchin, 1983). Other authors have in the two-tone (DA) condition than in the one-tone talked in terms of cortical control functions, rather than (FA) condition. A two-way ANOVA on these data in terms of processing resources. For example, Duncan's revealed a significant main effect for attention only, (1995) notion is that tasks, which show large drops in F(1,11)=21.22, p<.01. performance from full attention (FA) to DA conditions, require substantial amounts of controlled processing. By PET Data Duncan's view, episodic encoding tasks, which show a large performance drop under DA conditions, should Latent Variables and Task Contrasts require greater amounts of frontal control than episodic The Partial Least Squares (PLS) analysis produced three retrieval tasks, which show comparatively small drops LVs depicting the distribution of brain activity related under DA conditions. to particular task contrasts. The design scores for each Several previous studies have used PET in the DA latent variable (LV) are shown in Figure 1. The first LV (dual-task) paradigm, but to our knowledge, only two (LV1) distinguished FA from DA; the difference between have used episodic memory as one of the tasks (Fletcher, FA and DA design scores was larger during encoding Shallice, & Dolan, 1998; Shallice et al., 1994). These than during retrieval. Therefore, the singular image for investigators studied verbal learning while performing LV1 segregates brain activity related to DA from brain either an easy or a difficult perceptual-motor task. The activity related to FA, particularly during encoding. experiment thus examined the effects of DA during the The second LV (LV2) distinguished retrieval from encoding phase of a memory task; the effects of DA at encoding; the difference between retrieval and enco- retrieval were not studied. The major finding relevant to ding design scores was larger under conditions of FA the present experiment was thatÐrelative to the easy than DA. Therefore, the singular image for LV2 segre- distracting taskÐperformance of the difficult distracting gates brain activity related to retrieval from activity task greatly reduced activation in the left-prefrontal related to encoding, particularly under FA conditions. cortex. The last LV (LV3) revealed an interaction of the 268 Journal of Cognitive Neuroscience Volume 12, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdfplus/10.1162/089892900562093 by guest on 25 September 2021 Table 1. Behavioral Performance on the Memory Task and the ratio