Fig 1.
Pairs of BOS neurons and conditions for visual input and attention (modified from ref [39]).
A: Stimulus displays for testing the effects of object integration and selective attention. Ellipses on the borders of the keystone-like objects represent the receptive fields (RFs) of border ownership selective (BOS) neurons. In each display, three separate objects are presented. In the left display the RFs of two neurons are on the borders of two different objects (“unbound” condition). In the middle and right displays the two RFs lie on the borders of the same object (“bound” condition). Note that the visual stimuli in and around the RFs are identical in all three conditions, but represent parts of two separate objects in the left display and parts of the same object in the other two displays. In these experiments, the monkey attended one of the objects, as shown by a red star (not part of the display). Such an object is called “attended” while objects that are not attended are referred to as “ignored.” B: Consistent and inconsistent pairs. Arrows from an RF point toward the preferred side of the corresponding BOS neuron. Subscripts L and R indicate left and right side-of-figure preferences, respectively, while retinotopic position is represented by the superscripts “1” and “2”. RFs of neurons whose border ownership preferences are consistent with representing a common object are connected by black dashed lines (“consistent pairs”), while RFs of neurons with inconsistent preferences are connected by gray dashed lines (“inconsistent pairs”).
Fig 2.
Two different types of G-cells (balls with “G”) are part of the model: spatial grouping cells (Gsp) and object-based grouping cells (Gobj). Whereas Gsp-cells implement spatial attention, i.e. attention to everything within a circumscribed spatially defined area, Gobj-cells impart grouping structure of objects in the scene and mediate object-based attention. The latter are similar to the grouping cells from refs [25, 26, 41]. Feedback signals from these G-cells modulate activity of BOS neurons (balls with “BOS”) by NMDA-type connections (gray downward pointing arrows). Black and gray ellipses represent the locations of RFs of BOS neurons which are driven by visual input through AMPA-type synapses (black upwards pointing arrows). Black and gray horizontal arrows from RFs point toward the preferred side of the corresponding BOS neuron. For description of subscripts and superscripts of BOS cells see text.
Table 1.
Firing rates of model G-cells in different stimulus and attention conditions.
The labels in the first column refer to the stimulus and attention conditions in the RF of . Numbers in other columns are firing rates of G cells listed in the first row of the respective column, under conditions in the RF of
that are described in the first column.
Fig 3.
A: Raster plots showing 100 spike trains of BOS model neurons. For these plots, G-cells were activated for representing the Unbound-ignored condition between 0 and 1000 ms, the Bound-ignored condition for 1000 and 2000 ms, and the Bound-attended condition for 2000 and 3000 ms. Feedback from G-cells modulates the firing rates of BOS neurons. Identities of BOS neurons are shown next to each plot. B: Firing rates of preferred neurons ( and
). C: Firing rates of non-preferred neurons (
and
). The gray, black, and red bars represent the Unbound-ignored, Bound-ignored, and Bound-attended conditions, respectively, during trials that mimic the experimental settings of the Martin and von der Heydt study (ref. [39], their Fig 4B). These firing rates were obtained from 10 sets of 100 simulated trials, each of a length of 200 biological seconds. Error bars indicate standard deviations (SDs), which were very small in these simulations. Confidence intervals of preferred neurons were 9.42 ± 0.01 (SD = 0.02), 18.15 ± 0.01 (SD = 0.02), and 26.54 ± 0.01 (SD = 0.02) Hz for the Unbound-ignored, Bound-ignored, and Bound-attended conditions, respectively. Confidence intervals of non-preferred neurons were 18.15 ± 0.01 (SD = 0.02), 9.41 ± 0.01 (SD = 0.01), and 10.38 ± 0.01 (SD = 0.01) Hz for the Unbound-ignored, Bound-ignored, and Bound-attended conditions, respectively. Asterisks indicate significant differences between conditions (** p < 0.01 by t-test).
Fig 4.
Correlations between BOS neurons.
The gray, black, and red lines/bars represent the correlation (loose synchrony) of the Unbound-ignored, Bound-ignored, and Bound-attended conditions, respectively. A, C: Experimentally observed mean spike train cross-correlation and loose synchrony for consistent BOS neurons, modified from [39]. B, D: Model spike train cross-correlation and loose synchrony for the consistent pair. Confidence intervals of loose synchrony of this pair for the Unbound-ignored, Bound-ignored, and Bound-attended (panel D) were 1.13 ± 0.01 (SD = 0.01), 1.39 ± 0.02 (SD = 0.03), and 1.21 ± 0.01 (SD = 0.02) coincidences/s, respectively. E, G: Experimentally observed mean spike train cross-correlation and loose synchrony for inconsistent BOS neurons. F, H: Model spike train cross-correlation and loose synchrony for the inconsistent pairs. Inset in F shows detail of center region at higher scale. Confidence intervals of loose synchrony of these pairs for the Unbound-ignored, Bound-ignored, and Bound-attended (panel H) were 0.30 ± 0.01 (SD = 0.02), 0.23 ± 0.01 (SD = 0.01), and 0.34 ± 0.01 (SD = 0.01) coincidences/s, respectively. Curves in all panels are normalized by the maximum correlation value of the consistent pair. The observed maximum values were: A:54, B:24, E: 26, F:5 coincidences/s2. Asterisks indicate significant differences between conditions (** p < 0.01, t-test). Error bars indicate SDs.
Fig 5.
Tight synchrony between model BOS neurons.
Reduced cross-correlations after subtraction of Δ = 20ms interval jitter cross-correlation. A, C: Experimentally observed mean tight correlation and tight synchrony of the consistent pair for the Unbound-ignored (gray), Bound-ignored (black), and Bound-attended (red) conditions from [39]. B, D: Simulated tight correlation and tight synchrony of the consistent pair. Conventions are the same as those in A and C. Confidence intervals of tight synchrony of this pair for the Unbound-ignored, Bound-ignored, and Bound-attended conditions (panel D) were 0.0028 ± 0.0008 (SD = 0.0013), 0.0061 ± 0.0013 (SD = 0.0020), and 0.0076 ± 0.0018 (SD = 0.0029) coincidences/s, respectively. E, G: Experimentally observed mean tight correlation and tight synchrony of the inconsistent pair. F, H: Simulated tight correlation and tight synchrony of the inconsistent pairs. Inset in the panel F shows central area at higher scale. Confidence intervals of tight synchrony of these pairs for the Unbound-ignored, Bound-ignored, and Bound-attended conditions (panel H) were 0.0005 ± 0.0005 (SD = 0.0008), 0.0009 ± 0.0006 (SD = 0.0010), and 0.0012 ± 0.0005 (SD = 0.0007) coincidences/s, respectively. Curves in all panels are normalized by the maximum value of the “Bound-attended” condition of the consistent pair. The observed maximal values were: A:17.4, B:3.7, C:3.4, D:0.5 coincidences/s2. Asterisks indicate significant differences between conditions (* p < 0.05, ** p < 0.01, t-test). Error bars indicate SDs.
Fig 6.
Firing rates for BOS neurons as functions of the mean rates of G-cells.
A: Averaged firing rates of and
neurons for combinations of
- and Gsp-cells rates. B: Averaged firing rates of
and
neurons for combinations of
- and Gsp-cells rates. Firing rates of BOS neurons pointed out by white dashed lines with ‘C’, ‘D’, ‘E’ and ‘F’ are summarized in panels C, D, E and F as a functions of the mean rates of G-cells, respectively. In C-F, each data point is the average of 100 simulated trials, each 200 biological seconds long as a function of the mean rate of G-cells. C, D: Firing rates of BOS neuron as function of the mean rates of Gsp for two mean rates of
, 30Hz (filled triangles) and 60Hz (open triangles). C: Average of
and
. Black and red triangles indicate the chosen Gsp rates for ignored and attended conditions, respectively. D: Average of
and
neurons. Note that symbols for 30Hz and 60Hz nearly overlap because there is no direct projection from object-based G-cells to members of the inconsistent BOS cell pairs. E, F: Firing rates of BOS neurons as functions of the mean rates of
for two different mean rates of Gsp, 3Hz (filled triangles) and 15Hz (open triangles). E: Average of
and
. Black and red triangles indicate the chosen rates of
-cells for bound-ignored and bound-attended conditions, respectively. F: Average of
and
.
Fig 7.
Loose synchrony for model BOS neurons as function of the mean rates of G-cells.
Shown is the integral of the spike train cross-correlation in the range of a ±40 ms interval around lag zero. Each data point is the average of 100 simulated trials, each of 200 biological seconds duration. A: Loose synchrony between and
neurons (consistent pair) as a function of combinations of
- and Gsp-cells rates. B: Loose synchrony between all inconsistent pairs (defined in Fig 1B) of BOS neurons as function of combinations of
- and Gsp-cells rates. Modulation patterns of loose synchrony labelled by white dashed lines with ‘C’, ‘D’, ‘E’ and ‘F’ are summarized in panels C, D, E and F, respectively. C: Loose synchrony between consistent pair as a function of the mean rate of Gsp for two different mean rates of
, 30Hz (filled squares) and 60Hz (open squares). Black and red triangles are the same as those in Fig 6C. D: Loose synchrony between all inconsistent pairs (defined in Fig 1B) of BOS neurons as function of the rates of Gsp. Note that symbols of
for 30Hz and 60Hz nearly overlap because there is no direct projection from
-cells to members of the inconsistent BOS cell pairs. Conventions are the same as in panel C. E: Loose synchrony between consistent pair as function of the mean rates of
for different mean rates of Gsp, of 3Hz (filled squares and 15Hz (open squares). Black and red triangles are the same as those in Fig 6E. F: Loose synchrony between inconsistent pairs of BOS neurons as function of the rates of
. Conventions as in panel E.
Fig 8.
Tight synchrony for model BOS neurons as function of the mean rates of G-cells.
Shown is the integral of the tight correlation in the range of ±5 ms interval around lag zero. A: Tight synchrony between and
neurons (consistent pair) with respect to combinations of
- and Gsp-cells rates. B: Tight synchrony between all inconsistent pairs. Modulation patterns of tight synchrony labelled by white dashed lines with ‘C’, ‘D’, ‘E’ and ‘F’ are summarized in panels C, D, E and F as functions of the mean rates of G-cells, respectively. C: Tight synchrony for consistent pair as a function of the mean rates of Gsp for mean rates of
of 30Hz (filled circles) and 60Hz (open circles). Black and red triangles are the same as those in Fig 6C. D: Tight synchrony between BOS neurons for inconsistent pairs as a function of the Gsp rate. Conventions as in panel C. E: Tight synchrony between consistent pair as a function of the mean rates of
for mean rates of Gsp-cell of 3Hz (filled circles) and 15Hz (open circles). Black and red triangles are the same as those in Fig 6E. F: Tight synchrony between BOS neurons for inconsistent pairs as a function of
rate. Conventions as in panel F.
Fig 9.
Overlapping squares as an example of cluttered scenes.
In this illustration, the light-gray “foreground” square appears in front of the darker “background” rectangle; the perception is not that of a light-gray square adjacent to a darker L-shaped object. Red and black circles mark T and L junctions, respectively, and are not part of the visual display. Such local cues modulate the responses of BOS cells in addition to the grouping cell inputs discussed in the present study [25].