Shifting attentional priorities: control of spatial attention through hemispheric competition

Abstract

Regions of frontal and posterior parietal cortex are known to control the allocation of spatial attention across the visual field. However, the neural mechanisms underlying attentional control in the intact human brain remain unclear, with some studies supporting a hemispatial theory emphasizing a dominant function of the right hemisphere and others supporting an interhemispheric competition theory. We previously found neural evidence to support the latter account, in which topographically organized frontoparietal areas each generate a spatial bias, or "attentional weight," toward the contralateral hemifield, with the sum of the weights constituting the overall bias that can be exerted across visual space. Here, we used a multimodal approach consisting of functional magnetic resonance imaging (fMRI) of spatial attention signals, behavioral measures of spatial bias, and fMRI-guided single-pulse transcranial magnetic stimulation (TMS) to causally test this interhemispheric competition account. Across the group of fMRI subjects, we found substantial individual differences in the strengths of the frontoparietal attentional weights in each hemisphere, which predicted subjects' respective behavioral preferences when allocating spatial attention, as measured by a landmark task. Using TMS to interfere with attentional processing within specific topographic frontoparietal areas, we then demonstrated that the attentional weights of individual subjects, and thus their spatial attention behavior, could be predictably shifted toward one visual field or the other, depending on the site of interference. The results of our multimodal approach, combined with an emphasis on neural and behavioral individual differences, provide compelling evidence that spatial attention is controlled through competitive interactions between hemispheres rather than a dominant right hemisphere in the intact human brain.

Figure 1.

Landmark task and TMS target sites. A, Landmark task. Each block of the landmark task first began with instructions indicating whether subjects should judge which side of a horizontal line was longer or shorter. The same judgment was made for the entire block (example shows a “longer” block). Stimuli were horizontal lines shifted leftward or rightward in relation to a veridical midpoint defined by a vertical transection line. Note that the lines presented in the actual experiment were white, whereas the lines in this example are depicted as black for better contrast. Each trial started with a fixation point (displayed for 1500 ms), followed by a transected line stimulus (displayed for 200 ms), and ended with a full-screen mask (displayed for 2000 or 5000 ms during TMS). Subjects made their responses while the mask was displayed. During the TMS conditions, a single pulse (blue marker) was delivered over several ROIs (see below) at either 200 or 100 ms after the onset of the transected line stimulus during every trial. B, TMS target sites in PPC. Example in one subject (S3) of the three functional sites, right IPS1/2 (top row), right SPL1 (middle row), and left IPS1/2 (bottom row), that were targeted with TMS. Each topographic ROI is indicated by its own set of colored voxels. The green crosshairs indicate where the TMS coil was placed over each topographic area, as illustrated from a sagittal view (right column) and an axial view (left column).

Figure 2.

Spatial attention signals across human frontoparietal cortex. A, The neural bias toward the RVF or LVF, as assessed by an LI, for each individual topographic ROI averaged across the group of subjects (n = 12). Example of activations within frontal (top) and parietal (bottom) cortices projected onto an inflated surface of a representative subject's brain. Topographically organized ROIs are outlined in black. Numbers represent the average LI value for each topographic ROI. B, The LI values averaged across all ROIs within the LH and RH across subjects. Across the group, the average LI was not significantly different between hemispheres. Error bars indicate SEM. N.S., Not significant. C, The LI values averaged across all of the RH ROIs (x-axis) plotted against the LI values averaged across all of the LH ROIs (y-axis) for each individual subject (n = 12). There was a large amount of intersubject variability in LI values between hemispheres. Dashed line represents the equality line.

Figure 3.

Behavioral spatial biases. A, Examples of spatial biases obtained in two subjects (S12, left; S2, right) using the landmark task. Trials from each block were organized into bins based on their distance from the veridical midpoint (marked by the dashed line in each graph). The proportion reported as “right is longer” was then calculated for the set of trials that fell into a bin for a single block (thus, each × represents data from a single bin from 1 of the 4 blocks). The offset of the horizontal line to the left or the right was plotted against the proportion of trials reported as “right is longer,” and the data were fit with a Weibull function (A, black curves). Each individual subject's behavioral spatial bias was the point on the function in degrees of visual angle that corresponded to when left and right were reported equally often (i.e., the PSE; A, gray lines). Left, An example of a subject (S12) with a leftward behavioral bias (PSE is to the left of the veridical midpoint). Right, An example of a subject (S2) with a rightward behavioral bias (PSE is to the right of the veridical midpoint). B, Distribution of behavioral biases across all subjects tested in the landmark task (n = 45). The histogram represents the number of subjects with a given bias to the right (RVF bias) or the left (LVF bias) of the veridical midpoint (represented by the dotted line). For a subset of these subjects (n = 12), the behavioral bias values were correlated with their respective neural spatial bias (LI) values. The distribution of behavioral spatial bias values for these 12 subjects is shown in gray.

Figure 4.

Relationship between behavioral bias and neural bias. For each subject (n = 12), the behavioral spatial bias value (in degrees of visual angle), calculated using the landmark task, was plotted against the neural bias value (the average LI value across all topographically organized areas in parietal and frontal cortices), calculated from data gathered in a separate fMRI spatial attention experiment. The dotted line represents the regression line through the data. Subjects with stronger neural biases on average in the LH tended to show a bias toward the RVF while performing the landmark task, whereas those with stronger neural biases on average in the RH tended to show a bias toward the LVF while performing the landmark task.

Figure 5.

Behavioral spatial biases of individual subjects for each TMS condition. Examples of Weibull functions derived from data collected from four of the six subjects who underwent TMS while performing the landmark task. Each curve represents the data collected from a separate TMS condition (red curves, data from right SPL1 stimulation; blue curves, data from right IPS1/2 stimulation; green curves, data from left IPS1/2 stimulation; cyan curves, data from simultaneous left and right IPS1/2 stimulation). Each of the TMS conditions was compared with the data collected while subjects performed the landmark task without TMS (no-TMS condition; black curves). Solid curves represent data collected while TMS was applied 200 ms after stimulus onset, and the dashed curves represent data collected while TMS was applied 100 ms after stimulus onset. The vertical, solid black lines indicate the PSE/ behavioral spatial bias for each condition. All other conventions are as in Figure 3A.

Figure 6.

Behavioral spatial biases for each TMS condition averaged across subjects. The average behavioral spatial bias (in degrees of visual angle) across subjects (n = 6) for each TMS condition (red points, TMS over right SPL1; blue point, TMS over right IPS1/2; green point, TMS over left IPS1/2; cyan point, TMS over left and right IPS1/2) compared with the average behavioral spatial bias across subjects when TMS was not applied (no-TMS condition; black circle). Triangles indicate conditions during which TMS was applied 200 ms after stimulus onset. Diamond indicates the condition during which TMS was applied 100 ms after stimulus onset. *p < 0.05; **p < 0.01; N.S., not significant. Error bars indicate SEM.