Report

An Area Specialized for Spatial Working Memory in Human Frontal Cortex

See allHide authors and affiliations

Science  27 Feb 1998:
Vol. 279, Issue 5355, pp. 1347-1351
DOI: 10.1126/science.279.5355.1347

Abstract

Working memory is the process of maintaining an active representation of information so that it is available for use. In monkeys, a prefrontal cortical region important for spatial working memory lies in and around the principal sulcus, but in humans the location, and even the existence, of a region for spatial working memory is in dispute. By using functional magnetic resonance imaging in humans, an area in the superior frontal sulcus was identified that is specialized for spatial working memory. This area is located more superiorly and posteriorly in the human than in the monkey brain, which may explain why it was not recognized previously.

Studies of working memory in monkeys (1) and humans (2-7) have emphasized the important role of the prefrontal cortex. Physiological evidence for this role comes from studies demonstrating sustained activity in prefrontal cortex during working memory delays (1, 6-8). In monkeys, the dorsolateral prefrontal cortex within and surrounding the principal sulcus appears to be involved primarily in working memory for spatial locations, whereas the region ventral to it on the inferior convexity appears to be more involved in working memory for patterns, colors, objects, and faces (1, 9). The prefrontal region for spatial working memory in monkeys is located just anterior to a region for the control of eye movements, the frontal eye field (FEF), which is on the anterior bank of the arcuate sulcus (10).

Most functional brain imaging studies of spatial working memory in humans have focused on the dorsolateral frontal region defined by Brodmann as area 46 (5, 11, 12), because the spatial working memory region in monkeys lies within that area (13). While performance of spatial working memory tasks activates Brodmann area (BA) 46 (5, 11, 12), performance of working memory tasks involving other types of information, such as verbal and visual object information, does so as well [for example, see (3, 6, 7, 12, 14,15)]. Therefore, the existence of a prefrontal cortical area in humans that is specialized for spatial working memory has been questioned [for example, see (16)].

Here we provide evidence that a human frontal area specialized for spatial working memory does indeed exist, but it is not in BA 46. Rather than focusing on cytoarchitectonically defined BA 46, we focused our investigation on cortical areas in the vicinity of a functionally defined landmark that lies near the spatial working memory area in the monkey, the FEF. Imaging studies of eye movements have localized the human FEF to the precentral sulcus (17, 18). Because the monkey spatial working memory area lies just anterior to the FEF, we predicted that the human spatial working memory area might also lie just anterior to the FEF, perhaps within the superior frontal sulcus. Thus, we hypothesized that the location of this region for spatial working memory, like the human FEF, is relatively more superior and posterior than a similar functional region in monkey frontal cortex and has a different BA designation. Imaging studies of spatial working memory that have included dorsal and posterior frontal areas have consistently found activation in the vicinity of the superior frontal sulcus (5, 12, 15, 18-21), but a mnemonic role for this region has largely been dismissed because of its presumed location in premotor cortex or the FEF [but see (15, 18, 21)].

We investigated the role of the superior frontal sulcus in spatial working memory by using functional magnetic resonance imaging (fMRI). We used four criteria to determine whether this area is specialized for spatial working memory: (i) the area must show sustained activity during spatial working memory delays; (ii) such sustained activity must be greater during spatial working memory delays than during delays in other types of working memory tasks, in this case working memory for faces; (iii) the sustained activity during spatial working memory delays cannot be attributable to preparation for a motor response, which would indicate a premotor rather than a mnemonic function; and (iv) the area must be distinct from the FEF.

Functional MRI scans were obtained while 11 healthy human volunteers (8 males and 3 females) alternately performed individual trials of spatial working memory and sensorimotor control tasks (Fig.1). Activations associated with face working memory were also investigated in seven of these subjects, and those associated with visually guided saccades were investigated in the other four. Multiple regression analysis of the time course of the activation was used to identify voxels that were significantly activated by each task and to distinguish between activations that were due to different cognitive components of the tasks (7, 22). Activated voxels in frontal cortex were assigned to two dorsal regions of interest (the superior frontal sulcus and the precentral sulcus) and two ventral regions of interest (middle frontal cortex and inferior frontal cortex) based on anatomic criteria (23).

Figure 1

(left).(A) A face and a spatial location working memory task, which used the same stimuli and were equated for difficulty. Subjects saw a series of three faces, each presented for 2 s in a different location on the screen, followed by a 9-s memory delay. Then a single test face appeared in some location on the screen for 3 s, followed by a 6-s intertrial interval. Before each series, subjects were instructed to remember the locations or the identities of the three faces in the memory set. For the spatial task, the subject indicated with a left or right button press whether the test location was the same as one of the three locations presented in the memory set, regardless of which face marked that location. For the face memory task, the subject indicated whether the test face was the same as one of the three faces observed in the memory set, regardless of the location where the face appeared. For the sensorimotor control task, scrambled faces appeared (control stimulus set), and when the fourth scrambled picture appeared after the delay, subjects pressed both buttons (control response). For the control and working memory tasks, subjects were instructed to look directly at each picture as it appeared and to avoid moving their eyes during delay. All subjects gave written informed consent. fMRI time series data were analyzed by multiple regression (7, 22). Contrasts between task components are shown below the task diagram: (1) visual stimulation versus no visual stimulation; (2) memory stimuli versus control stimuli; (3) control stimulus set versus control response; (4) memory stimulus set versus test stimulus and response; (5) delays during anticipation of response versus intertrial intervals; (6) memory delays versus control delays. Contrasts were constructed a priori so the sum of values over the time series equals zero and the crossproducts of all pairs of contrasts equal zero, demonstrating orthogonality. These contrasts make it possible to obtain independent estimates of activity levels during every phase of the task. The specific contrasts were chosen to test the hypotheses that motivated this study. Each time series was convolved with a Gaussian model of the hemodynamic response to produce the six regressors used in the analysis. Multiple regression simultaneously calculates a weighting coefficient for each regressor so the sum of the regressors multiplied by their weighting coefficients provides the best fit to the data. (B and C) fMRI time series data (solid line) and the corresponding fitted response functions (dashed line) from regression analysis. (B) One voxel in the superior frontal sulcus showing sustained activity during the working memory delay (regression coefficients: 2.06*, 0.75*, 0.43, 0.32, 0.23*, 1.43*). (C) One voxel from the precentral sulcus showing only transient activity during stimulus presentation (regression coefficients: 3.56*, 0.79, −0.73, 1.14, 0.57*, −0.03). Asterisks mark regressors that account for a significant portion of the variance in that voxel independent of the variance attributable to the other regressors.

All subjects showed sustained activity during spatial working memory delays (regressor 6) (Fig. 1) in dorsal frontal cortex. Within dorsal frontal cortex, all subjects had a more extensive region of sustained activity in the superior frontal sulcus than in the precentral sulcus. Across subjects, 66% (median) of dorsal frontal cortex demonstrating sustained activity was in the superior frontal sulcus. By contrast, 69% of dorsal frontal cortex demonstrating only transient activity during the presentation of visual stimuli (regressor 1) was in the precentral sulcus, just posterior and lateral to the sustained activation in the superior frontal sulcus. This difference between the localizations of sustained and transient activity demonstrates two functionally distinct areas in the dorsal frontal cortex (P < 0.002 for response-type by sulcus interaction). Thus, the superior frontal sulcus meets our first criterion for a spatial working memory area by demonstrating sustained activity during delays.

To determine whether the region showing sustained activity in the superior frontal sulcus is specifically involved in spatial working memory rather than more generally in any working memory task, fMRI data were obtained (n = 7) during performance of a face working memory task that used identical stimuli, timing, and motor response, as did the spatial working memory task (Fig. 1). The superior frontal sulcus bilaterally showed significantly more sustained activity during spatial than during face working memory delays (5.4 cm3 versus 2.3 cm3 of cortex, 0.48% versus 0.37% mean signal change, respectively, summed across both hemispheres; P < 0.01 for both spatial extent and signal intensity change) (Fig. 2). By contrast, left inferior frontal cortex showed significantly more sustained activity during face than during spatial working memory delays (4.9 cm3 versus 2.7 cm3, 0.44% versus 0.21%, for left inferior frontal cortex; P < 0.05 for both comparisons). The left middle frontal cortex showed significantly more cortical area activated by face working memory than by spatial working memory (8.9 cm3 versus 5.7 cm3;P < 0.05), but the difference in activation based on the change in mean signal intensity failed to reach significance (0.36% versus 0.22%; P > 0.1). The difference between face and spatial working memory in the right ventral frontal cortex failed to reach significance for the middle or the inferior regions (24). There were no hemisphere effects for spatial working memory (P > 0.1 for superior, middle, and inferior frontal regions) (25). This comparison of spatial and face working memory demonstrates a double dissociation (region by task interaction; P < 0.01 for both spatial extent and signal intensity change) with the superior frontal sulcus showing more sustained activity during spatial working memory delays and with the left inferior frontal region showing more sustained activity during face working memory delays. This comparison also demonstrates that the sustained activity in the superior frontal sulcus cannot be attributed to preparation for a motor response because the spatial and face working memory tasks required the same types of responses after delays. In short, the results of this comparison show that the superior frontal sulcus meets our second and third criteria for a spatial working memory area.

Figure 2

(right).Areas with significant sustained activity in a single subject during the working memory delay for faces (blue outline) and for spatial locations (red outline) overlaid onto the subject's Talairach normalized anatomical MR image. Level above the bicommissural plane is indicated for each axial section. Twenty-one contiguous axial slices were obtained in series of 88 scans each (repeat time, 3 s; echo time, 40 ms; flip angle 90°). Slices were either 5 or 6 mm thick as needed to cover the entire brain. L, left; R, right.

To determine whether this region in the superior frontal sulcus is distinct from the FEF, we compared the locations of activations during the spatial working memory task with activations evoked during a saccadic eye movement task in the same scanning session (n = 4) (Fig. 3). As expected from previous results (17, 18), most of the dorsal frontal cortex activated during the eye movement task (85% median across subjects) was in the precentral sulcus, thereby identifying the location of the FEF in each subject and distinguishing it from the region of sustained activity in the superior frontal sulcus (Fig. 3). In three of four subjects, the centroid of sustained activity was significantly anterior to the centroid of eye movement–related activity in both right and left hemispheres (mean difference = 8.5 mm; P < 0.005 in all cases). The fourth subject showed the same trend, but the difference (3.5 mm) did not reach statistical significance because the areas of sustained activity were less extensive than in the other subjects. In addition, the face task was designed to require the same types of eye movements as the spatial task and yet showed significantly less sustained activity in the superior frontal sulcus, further supporting our assertion that the sustained activity there is related to spatial working memory and not to oculomotor control. Thus, the region of sustained activity in the superior frontal sulcus is distinct from the FEF, thereby meeting our fourth criterion for a spatial working memory area (26).

Figure 3

(Top) Results from a single subject for the spatial working memory task and the saccadic eye movement task. For the eye movement task, subjects made a random series of horizontal visually guided saccades with an average amplitude of 12° (range, 5° to 20°). Subjects performed saccades during 15-s blocks interleaved with 15-s blocks in which subjects saw a blank screen and were instructed to look at the center of the screen and avoid moving their eyes. Yellow outline indicates the region activated by the saccadic eye movement task. Green outline indicates the region activated transiently during the presentation of the stimuli in the working memory task. Red outline indicates the region that showed sustained activation during the working memory delay. (Bottom) White dotted line marks superior frontal sulcus. Yellow dotted line marks precentral sulcus. Mean locations of the local maximum Z-scores for saccadic eye movements (yellow) and for transient (green) and sustained (red) activity during the spatial working memory task are shown on the average of the Talairach normalized anatomical MRI scans from the four subjects who performed both the spatial working memory and the saccadic eye movement tasks. Center and radii of the cylinders indicate mean and standard deviations of the locations of local maxima, respectively. Saccades: left, −33 ± 8 , −16 ± 9, +46 ± 5; right, +29 ± 8, −9 ± 4, +45 ± 5. Transient: −35 ± 6 , −17 ± 7, +45 ± 6; +30 ± 8, −8 ± 7, +46 ± 4. Sustained: −31 ± 7, −7 ± 5, +46 ± 4; +27 ± 5, −5 ± 4, +49 ± 5. Note that the sustained spatial working memory activation is just anterior to the transient and saccade activations. L, left; R, right. VAC and VPC, vertical planes passing through the anterior and posterior commissures, respectively.

The results of the current fMRI study indicate that there is a functional area in the superior frontal sulcus that shows sustained activity during spatial working memory delays. Sustained activity in this region was greater during spatial than during face working memory delays, demonstrating that the superior frontal sulcus plays a predominant role in spatial working memory compared with object working memory. Because these two tasks were matched for motor preparation and motor response, sustained activity in the superior frontal sulcus cannot be attributed to manual motor function. Moreover, this area is distinct from, and just anterior to, the human FEF, demonstrating a correspondence to the topological relationship that has been described in the monkey between a frontal cortical area for spatial working memory and the FEF [for example, see (27)]. However, in the monkey the frontal cortical area for spatial working memory is within the principal sulcus and surrounding cortex just anterior to the FEF, which is on the anterior bank of the arcuate sulcus. Our data indicate that the homologous functional areas in the human brain are in the superior frontal and precentral sulci, respectively, and thus occupy a relatively more superior and posterior location. The location of this spatial working memory area in humans, superior and posterior to BA 46, may explain why it was not recognized in previous studies.

This difference in functional neuroanatomy between monkeys and humans suggests that the spatial working memory area and the FEF in humans have been displaced by the expansion of the more inferior and anterior portions of the lateral prefrontal cortex over the course of primate brain evolution. Comparison of monkey and human functional neuroanatomy indicates that displacement of regions in the human brain may be due to the emergence of phylogenetically newer regions. For example, extrastriate visual areas specialized for spatial vision have a more superior location in parietal cortex in the human than in the monkey, whereas those specialized for object vision have a more inferior location in temporal cortex [for review, see (28)]. Displacement of both sets of visual areas away from the posterior perisylvian cortex may be related to the emergence of language mediated by phylogenetically newer cortical areas such as BA 39 and BA 40. The specific displacement of functional areas in dorsal frontal cortex that we have identified may likewise be related to the emergence of other cognitive abilities, either distinctively human or greatly elaborated in humans, mediated by new functional areas in prefrontal cortex. Examples of these abilities may include abstract reasoning, complex problem solving, and planning for the future, consistent with behavioral symptoms in patients with frontal lobe damage (29).

  • * To whom correspondence should be addressed. E-mail: Susan Courtney{at}nih.gov

REFERENCES AND NOTES

View Abstract

Navigate This Article