Fig 1.
Experimental task and behavior.
(A) Observers saw a brief sample display (250 ms) that contained 1 or 2 colored dots. After a delay period (1,000 ms), the fixation point changed color (blue or green) to indicate which item should be reported. Observers reported the angular position of the cued item as precisely as possible by mouse click on the perimeter of a rim. (B) Median response time (ms) as a function of memory set size (1 versus 2 items). Light grey lines represent individual observers. Black lines represent the mean. Error bars represent ±1 SEM. (C–E) Parameter estimates obtained by fitting a three-component mixture model [50] to response errors as a function of memory set size. s.d. (panel C) reflects precision of responses (with higher values indicating worse precision), pGuess (panel D) estimates the probability that the observer produced a random response (i.e., a guess), and pSwap (panel E) estimates the probability that the uncued item was misreported instead of the cued item. Note that pSwap is necessarily 0 for the single-item condition. Data are available at https://osf.io/47cmn/. s.d., standard deviation of von Mises distribution.
Fig 2.
IEM procedure for reconstruction spatial CTFs.
(A) The stimuli varied in angular position around the fixation point. Stimuli were categorized as belonging to 1 of 8 positions bins (a 45° wedge of positions), centered at 0°, 45°, and so forth. (B) We modeled power at each electrode as a linear combination of the 8 hypothetical spatial channels (C1–C8). The curves show the predicted response of the spatial channels across angular positions. (C) In the training phase of the analysis, we estimated the relative contributions of each of the 8 channels to power measured at each electrode (called “channel weights”). (D) In the test phase of the analysis, we used the channel weights obtained during training to estimate the channel responses from the multivariate pattern of power across electrodes. Critically, the data used in the training and test phases of the analysis were independent to avoid circularity. For details, see Materials and methods. Data are available at https://osf.io/47cmn/. CTF, channel-tuning function; IEM, inverted encoding model.
Fig 3.
Spatial alpha-band CTFs as a function of memory load.
Average alpha-band CTF in the one- and two-item conditions (A and B, respectively). (C) The spatial selectivity of alpha-band CTFs across time (measured as CTF slope, see Materials and methods) as a function of memory load. The blue (one-item) and red (two-item) markers at the top of the panel indicate the period of above-chance selectivity obtained using a cluster-based test. CTF selectivity was reliably higher in the one-item condition than in the two-item condition. The shaded error bars reflect ±1 SEM across observers. Data are available at https://osf.io/47cmn/. CTF, channel-tuning function.
Fig 4.
Testing whether multivariate patterns of alpha-band power generalize between one-item and two-item trials.
(A) The spatial selectivity of alpha-band CTFs across time (measured as CTF slope; see Materials and methods) in the two-item condition when the IEM was estimated using the one-item condition. The markers at the top of the panel indicate the period of above-chance selectivity obtained using a cluster-based test. (B) Mean delay-period (averaged from 250–1,250 ms) channel-response profiles as a function of the distance between the items. The arrows above each subplot mark the channel(s) tuned for the positions of the 2 stimuli. The shaded error bars reflect ±1 SEM across observers. Data are available at https://osf.io/47cmn/. CTF, channel-tuning function; IEM, inverted encoding model.
Fig 5.
Alpha-band activity concurrently represents 2 spatial positions.
(A) If only 1 item can be represented by alpha-band activity at a time, then in the two-item condition, the items might alternate between an active and an activity-silent state such that only 1 item is actively represented at once. We used data from the single-item condition to simulate the expected CTF based on this switching account. This account holds that each item is represented 50% of the time (on average). Thus, we simulated switching between items by randomizing the position labels for 50% of trials for the one-item condition. (B) Spatial selectivity of alpha-band CTFs across time (measured as CTF slope) for the two-item condition (red) and for simulated switching (black). CTF selectivity was reliably higher during the two-item condition (red) than for simulated switching (black). The shaded error bars reflect ±1 SEM across observers. Data are available at https://osf.io/47cmn/. CTF, channel-tuning function.
Fig 6.
Identifying frequencies that track remembered positions during one-item and two-item maintenance.
(A–B) Selectivity of spatial CTFs (measured as CTF slope) reconstructed from the scalp distribution of oscillatory power across a broad range of frequencies for the one-item (A) and two-item (B) conditions. Points with no reliable CTF selectivity as determined by the cluster-corrected permutation test are set to dark blue. (C) Overlay plot marking the clusters of reliable selectivity in the one-item condition (light blue), two-item condition (red), and both (orange). Data are available at https://osf.io/47cmn/. CTF, channel-tuning function.