A Possible Neural Basis for Attentional Capture of Faces Revealed by Functional Magnetic Resonance Imaging and Causal Pharmacological Inactivation in Macaques

Abstract

In primates, the presence of a face in a visual scene captures attention and rapidly directs the observer’s gaze to the face, even when the face is not relevant to the task at hand. Here, we explored a neural circuit that might potentially play a causal role in this powerful behavior. In our previous research, two monkeys received microinfusions of muscimol, a γ-aminobutyric acid type A (GABA_A)-receptor agonist, or saline (as a control condition) in separate sessions into individual or pairs of four inferotemporal face patches (middle and anterior lateral and fundal), as identified by an initial localizer experiment. Then, using fMRI, we measured the impact of each inactivation condition on responses in the other face patches relative to the control condition. In this study, we used the same method and measured the impact of each inactivation condition on responses in the FEF and the lateral intraparietal area, two regions associated with attentional processing, while face and nonface object stimuli were viewed. Our results revealed potential relationships between inferotemporal face patches and these two attention-related regions: The inactivation of the middle lateral and anterior fundal face patches had a pronounced impact on FEF, whereas the inactivation of the middle and anterior lateral face patches had a noticeable influence on LIP. Together, these initial exploratory findings document a circuit that potentially underlies the attentional capture of faces. Confirmation of the role of this circuit remains to be accomplished in the context of a paradigm that explicitly tests the attentional capture of faces.

INTRODUCTION

Out of the myriad of objects encountered in daily life, faces are the most socially relevant; they convey information not only about age, sex, and identity but also about the gaze direction, intention, and emotional response of the viewed individual. Numerous empirical studies have confirmed that human observers are drawn to attend to faces more rapidly than to other objects (Morrisey, Hofrichter, & Rutherford, 2019; Hodsoll, Viding, & Lavie, 2011; Davies & Hoffman, 2002). This bias, or “attentional capture,” is especially evident when humans view human faces but is also observed when they view animal faces (Lin et al., 2024; Jakobsen, White, & Simpson, 2021).

A particular compelling demonstration of this attentional bias occurs when human observers search for a nonface target in a scene: Their performance is significantly slowed when a task-irrelevant face is present compared with when a nonface stimulus is present, and this slowing is eliminated when face processing is hindered, for example, when the face image is inverted (Jenkins & Langton, 2003; see also Sato & Kawahara, 2015). Consistent with the bias toward faces in the input, target detection is enhanced when a probe (a small grey square) appears in a location previously occupied by a face rather than in a location occupied by an object (Bindemann, Burton, Langton, Schweinberger, & Doherty, 2007). Furthermore, the rapid attention to faces is reflected in the capture of first saccades by faces, with the duration of ensuing fixation within the face stimuli being shorter than for other stimuli, in line with “ultra-fast” face perception (Morrisey et al., 2019). This advantage for face over nonface processing is compatible with the presence of a widespread face circuit in humans (Avidan & Behrmann, 2009, 2021; Gobbini & Haxby, 2007; Haxby, Hoffman, & Gobbini, 2000, 2002). Within this circuit, the fusiform face area, occipital face area, and STS comprise the “core” regions, whereas, in humans, the anterior temporal cortex, precuneus, anterior paracingulate cortex, inferior frontal gyrus, amygdala, insular, and reward system comprise the “extended” areas. Despite the thorough characterization of these key face areas in humans and their multiple face-patch homologues in nonhuman primates (NHPs) (Hori et al., 2021; Bell, Hadj-Bouziane, Frihauf, Tootell, & Ungerleider, 2009; Tsao, Moeller, & Freiwald, 2008; Tsao, Freiwald, Knutsen, Mandeville, & Tootell, 2003), the cortical substrate that specifically subserves the attentional capture of faces has not been investigated. Notably, cortical regions associated with attentional capture of faces are not included in circuit diagrams of the distributed face network.

Given that, intuitively, the attentional capture of faces probably requires the engagement of one or more face-selective regions communicating with one or more areas associated with visual attentional processing, here, we set out to identify attention-related regions that might be implicated. Two candidate cortical regions that might serve this role and might interact with face-selective regions include the FEF and the lateral intraparietal area (LIP) (Kastner & Ungerleider, 2000), both of which have been well characterized in humans and NHPs. These two regions are activated under conditions of covert as well as overt (eye movements) attention (Panichello & Buschman, 2021; Fiebelkorn & Kastner, 2020; Corbetta & Shulman, 2002; Corbetta, 1998), and their cooperation appears to be tightly coordinated via a push–pull mechanism (Hall, Colby, & Olson, 2020). As such, both FEF and LIP may be strong contenders for enabling interactions between face perception and attentional engagement in the context of attentional capture and rapid fixations to faces. In the present study, we aimed to explore an underlying mechanism that might support the rapid attention to faces and to provide some initial data that point to a potential neural correlate of the capture by a face in the visual input in NHPs.

Adopting NHPs as a model for studying the attentional capture of faces depends on the longstanding assumption of homologies between humans and monkeys for both face recognition (Zaldivar et al., 2022; Yovel & Freiwald, 2013; Leopold & Rhodes, 2010; Tsao, Moeller, et al., 2008; Gobbini & Haxby, 2007) and attention systems (Kastner & Ungerleider, 2008). Indeed, much evidence favors these cross-species similarities (but see Rossion & Taubert, 2019). Just as in humans, in monkeys, there is also a large swath of core regions, composed of at least six inferotemporal (IT) patches distributed along the STS, which are disproportionately activated by faces (Chang, Egger, Vetter, & Tsao, 2021; Tsao & Livingstone, 2008; Tsao, Moeller, et al., 2008; Tsao et al., 2003). Likewise, as in humans, in monkeys, there are additional extended face-responsive regions, including the amygdala (Liu, Hadj-Bouziane, Moran, Ungerleider, & Ishai, 2017; Liu et al., 2015; Hadj-Bouziane et al., 2012), frontal cortex (Tsao, Schweers, Moeller, & Freiwald, 2008), and anatomically conserved areas in the temporal pole as well as perirhinal cortex whose nonlinear activations are associated with familiar over unfamiliar faces (Landi & Freiwald, 2017).

Critically, as is the case for humans, faces are highly salient for NHPs (Mosher, Zimmerman, & Gothard, 2014) and the presence of a face in the visual input also results in attentional capture. For example, monkeys’ monitoring of a color change in a peripheral target is impeded in the presence of a task-irrelevant face, especially one with an emotional expression (Landman, Sharma, Sur, & Desimone, 2014), for example, when the teeth of the face stimulus are visible in a fear grimace (Carp et al., 2022). Monkeys also make rapid saccades within 100–110 msec to faces, considerably faster than saccades to images of animals or vehicles, and they show preferential attentional search for faces versus other stimuli (Crouzet, Kirchner, & Thorpe, 2010). Relatedly, chimps and bonobos rapidly attend to conspecific faces, especially those with whom they have a positive relationship, even after decades of separation, suggesting that information about social relationships may strongly influence the biased attention to specific faces (Lewis et al., 2023). Notably, recent studies in monkeys found that the inactivation of middle face patches impairs face detection performance and alters eye movements during free viewing of faces (Azadi, Lopez, Taubert, Patterson, & Afraz, 2024; Sadagopan, Zarco, & Freiwald, 2017).

As in humans, the response properties of both FEF and LIP in NHPs make them candidate regions for a role in attentional capture of faces. For example, when a visuospatial target is attended, there is an increase in neural activity in monkey FEF, independent of its oculomotor function (Soyuhos & Baldauf, 2023; Armstrong, Chang, & Moore, 2009) or saccade execution (Thompson, Biscoe, & Sato, 2005). The role of FEF in spatial attention has also been confirmed in inactivation studies and studies using suprathreshold microstimulation (Moeller, Freiwald, & Tsao, 2008; Moore & Armstrong, 2003), and FEF interacts dynamically with extrastriate visual cortex (Veniero et al., 2021) directly or indirectly through connectivity with the intraparietal sulcus and LIP (Merrikhi et al., 2017; Gregoriou, Gotts, & Desimone, 2012; for recent review, see Martinez-Trujillo, 2022, indicating the possible link with face processing patches). LIP, on the other hand, plays a role in active exploration via eye movements and has direct projections to oculomotor centers (Brunamonti & Pare, 2023). In addition, LIP is engaged in the control of attentional deployment in space (Bisley & Goldberg, 2003), and inactivation of this region impedes visual search and target detection in the contralateral field (Wardak, Olivier, & Duhamel, 2004).

To interrogate the potential neural correlates of attentional capture of faces, we built on our previous investigation of the core face circuit (Liu et al., 2022), which used the same methodology adopted herein. In this previous study, to uncover the causal relationships among four face patches (two face patches near the junction between the occipital and temporal lobes [TEO], one on the middle fundal [MF] and one on the middle lateral [ML] surface, and two face patches near area TE, one on the anterior fundal [AF] and one on the anterior lateral [AL] surface), we injected, in separate sessions, either muscimol or a control, saline, into each patch (or a pair of patches) and, using fMRI, measured alterations in evoked neural response to faces or objects in the remaining patches. We uncovered a dual route (dorsal-ventral) fundal-lateral functional organization in the IT face network. This study enabled us to decompose the core face circuit into multiple functional compartments and to determine their directional signal propagation. Importantly, this previous study did not evaluate the coupling of face regions with cortical attention-related areas.

The data reported in the present article were collected from the same two monkeys as in the previous study. Here, using the identical perturbation (transient inactivation) and measurement (fMRI) approach, we documented the consequences of inactivating an individual face patch, or a pair of IT patches, on the face versus nonface object responses in two key attentional areas, LIP and FEF. If FEF and/or LIP play a functional role in the attentional capture of faces, then the inactivation of a face patch(es) should result in a reduction in responses to faces and perhaps a weaker or reduced response to objects. Inactivation of anterior versus middle face patches and its consequences on FEF and LIP responses would reveal connectivity between the IT face patch(es) and attention-related areas that may be necessary for the attentional capture of faces. It is critical to note that, although we did not conduct an attentional capture study per se, our goal was to pinpoint an interactive circuit so that future studies can target this face-attention circuit using a paradigm that explicitly evokes the attentional capture of faces.

METHODS

Here, we briefly present the key aspects of the methods. Please see additional details in our previous study (Liu et al., 2022).

Participants and General Procedures

Two male macaque monkeys participated in these experiments (Monkeys C and D, Macaca mulatta, 9 years old; 6.5–7.5 kg). All procedures followed the Institute of Laboratory Animal Research (part of the National Research Council of the National Academy of Sciences) guidelines and were approved by the NIMH Animal Care and Use Committee. Each monkey was surgically implanted with a magnetic-resonance-compatible head post. Before surgery, the animal received atropine (0.05 mg/kg, im) and was lightly anesthetized with ketamine (10–20 mg/kg, im) before intubation. Then, the animal was given isoflurane (1.5–3%, to effect) as a general anesthetic. The administration of these medications was done by highly trained staff under the supervision of a veterinarian. Throughout the procedure, the animal’s body temperature was carefully regulated using a heating pad, and their heart rate, temperature, and respiration were continuously monitored. All surgical procedures were performed using aseptic techniques to minimize the risk of infection. Cefazolin (15 mg/kg, im) was administered the day before surgery and for a week after the surgery as a prophylactic against infection; for analgesia, ketoprofen (2.2 mg/kg, im) was administered immediately after the surgery and for the next 3–5 days. After recovery, the monkeys were trained to sit in a plastic chair and fixate on a central target presented on a screen for long durations with a stable head position (Liu et al., 2013, 2015, 2022).

Brain Activity Measurements

fMRI and anatomical MRI scanning was carried out in the Neurophysiology Imaging Facility Core (NIMH, National Institute of Neurological Disorders and Stroke, National Eye Institute [NEI]). Before each scanning session, a contrast agent (monocrystalline iron oxide nanocolloid [MION]) was injected into the femoral or external saphenous vein (12–15 mg/kg) to increase the contrast/noise ratio and to optimize the localization of fMRI signals (Leite et al., 2002). Imaging data were collected in a 4.7 T Bruker scanner with a surface coil array (two elements in Monkey C and eight elements in Monkey D). Please see the detailed scanning parameters in our previous study (Liu et al., 2022).

Experimental Design and Task

To identify ROIs, we performed an initial localizer experiment (Liu et al., 2015; Liu et al., 2013). The stimuli were presented in a block design. For Monkey C, grayscale photos of neutral monkey faces, familiar places, and familiar objects, along with their corresponding Fourier-phase versions, were presented in separate blocks. Each experimental block lasted 30 sec and was presented once in each run. For Monkey D, the same set of intact images and Fourier-phase scrambled faces used for Monkey C were presented in separate blocks. Each block lasted 32 sec and was presented twice in each run. Note that the same set of images was presented consistently to the animals before the experiments (e.g., during training and testing scanning) and throughout the experiments. As such, both images of unfamiliar monkeys and images of familiar objects inevitably became familiar to the monkeys and, therefore, any potential impact of stimulus familiarity is likely to be minimal in our study.

In the subsequent inactivation and corresponding control imaging experiments, to optimize the statistical power, only grayscale photos of neutral monkey faces and familiar objects were presented to the animals (Figure 1A for example stimuli). Each categorical block lasted 32 sec and was presented 4 times in each run. In all experiments, each categorical block alternated with 20-sec fixation blocks. Individual runs began and ended with a fixation block. Different pseudorandom sequences were used in each run. In each categorical block, 16 images were each presented for 700 msec followed by a 300-msec interval and repeated twice. To encourage prolonged fixation, a classic reward paradigm in fMRI of awake monkeys was used where rewards increased in a staircase style with longer fixation durations (Hadj-Bouziane et al., 2012; Bell et al., 2009). Notably, no discernible patterns were observed in the occurrence of fixation breaks, which randomly occurred both within the category blocks and during the baseline period. Therefore, the influence of this reward schedule on the current results should be minimal. Rewards were controlled by a QNX system. Data were included from only those runs in which fixation was maintained on at least 90% of the runs. In the remaining runs, no categorical blocks were excluded from the analysis. The stimuli were presented using the Presentation software (Version 12.2, www.neurobs.com).

Figure 1.

The fMRI experimental design and face-selective patches in Monkey C. (A) An illustrative example of the fMRI experimental design in inactivation and corresponding control experiments. (B) face-responsive (neutral monkey faces vs. scrambled faces, p < .005 uncorrected) activation maps from the initial localizer sessions are shown on lateral views of the right inflated cortex. The face-selected ROIs are encircled by black lines. FEF and LIP ROIs are encircled by green lines.

ROIs

IT Face-selective ROIs

As in our previous study, a two-step ROI definition method was employed for the definition of face patches, taking into account the impact of MION accumulation. First, all runs from the initial localizer were concatenated. Each monkey was scanned in two to three localizer sessions, resulting in multiple runs for each. For each monkey, we identified IT face patches using the contrast of neutral monkey faces versus objects (p < .005 uncorrected; false-discovery-rate-corrected q value is 0.0056 for Monkey C and 0.0209 for Monkey D).

Attention-related ROIs

Next, we identified the peak of activation around FEF and LIP using a contrast of neutral monkey faces versus scrambled faces (p < .005 uncorrected; false-discovery-rate-corrected q value is 0.0031 for Monkey C and 0.0438 for Monkey D). These “face-responsive” voxels within a radius of 3 mm around the peak voxel in the FEF and LIP were combined to yield the FEF and LIP ROIs, respectively.

We also conducted a more specific analysis. Because FEF and LIP are known to have voxels that are face-selective (Rajimehra, Young, & Tootell, 2009), the IT face patch inactivation may have effects on face-selective neurons in FEF and LIP, preventing us from understanding the relationship between face-selective responses and attention-related responses. Thus, we redid the analyses by identifying and then removing any “face-selective” (faces minus objects) voxels (p < .05 uncorrected) in FEF and LIP ROIs. For the FEF ROIs, 18 and 10 voxels were excluded from the initial pool of 54 face-responsive voxels in Monkey C and 35 face-responsive voxels in Monkey D, respectively. For the LIP ROIs, 27 and 2 voxels were excluded from the initial pool of 63 face-responsive voxels in Monkey C and 23 face-responsive voxels in Monkey D, respectively.

Because of the MION accumulation, the signals were weaker in this experimental portion compared with the data collected earlier in the experiments (e.g., the initial localizer). Including voxels with such weak responses in the control condition would hamper the comparisons between the control and inactivation conditions. Hence, a second step for ROI definition was conducted (see details in Liu et al., 2022). We concatenated all the control sessions under the same set of inactivations and then identified the face-selective voxels (faces vs. objects, p < .005 uncorrected for the IT cortex) as well as the face-responsive voxels (faces vs. baseline, p < .001 uncorrected). Any voxels within the initially defined ROIs that could not be identified in the corresponding control sessions were removed to yield the final ROIs for each type of inactivation sessions. Furthermore, in the present study, for a given control session (saline injection), if no voxel survived the initially defined FEF/ LIP ROIs, it and subsequent session(s) within the same set of inactivation were excluded from the subsequent analyses.

Given the above restrictions, the number of sessions per injection site varied for FEF and LIP (see details in Table 1). Briefly, for the inactivation of the middle face patches, we collected 16–17 control sessions from both animals, encompassing 133–143 runs for LIP and FEF, respectively. In addition, there were four sessions for the combined MF and ML inactivation (MF&ML) inactivation (the number of runs = 32), five sessions for the MF inactivation (the number of runs = 45), and 10 sessions for the ML inactivation (FEF: the number of runs = 82; LIP: the number of runs = 85). Regarding the inactivation of the anterior face patches, we gathered eight to nine control sessions from both animals, consisting of 69–74 runs for FEF and LIP, respectively. Moreover, there were two to three sessions for the combined AF and AL inactivation (AF&AL inactivation; FEF: the number of runs = 20; LIP: the number of runs = 27), three to five sessions for the MF inactivation (FEF: the number of runs = 27; LIP: the number of runs = 43), and one to three sessions for the ML inactivation (FEF: the number of runs = 8; LIP: the number of runs = 26). A large number of runs gathered from two individuals (as is often the case in NHP studies) is similar in design to the approach adopted in psychophysics studies, which includes a large number of trials in a few participants, and similar to fMRI studies, which permit a deep analysis of a single individual’s data (Gordon et al., 2023; Kriegeskorte, Formisano, Sorger, & Goebel, 2007).

Table 1.

Summary of the Different Types of Face Patch Inactivations.

	Monkey C	Monkey C	Monkey D	Monkey D
	LHem	RHem	LHem	RHem	Total
FEF
M-Control	50 (6)	26 (3)	25 (3)	42 (5)	143 (17)
MF&ML inactivation		16 (2)	16 (2)		32 (4)
MF inactivation		10 (1)	35 (4)		45 (5)
ML inactivation	42 (5)	14 (2)		26 (3)	82 (10)
LIP
M-Control	34 (4)	26 (3)	23 (3)	50 (6)	133 (16)
MF&ML inactivation		16 (2)	16 (2)		32 (4)
MF inactivation		10 (1)	35 (4)		45 (5)
ML inactivation	37 (4)	14 (2)	8 (1)	26 (3)	85 (10)
FEF
A-Control	18 (2)	19 (2)		32 (4)	69 (8)
AF&AL inactivation	10 (1)	10 (1)			20 (2)
AF inactivation	27 (3)				27 (3)
AL inactivation				8 (1)	8 (1)
LIP
A-Control	18 (2)		24 (3)	32 (4)	74 (9)
AF&AL inactivation	10 (1)		17 (2)		27 (3)
AF inactivation	27 (3)		16 (2)		43 (5)
AL inactivation			18 (2)	8 (1)	26 (3)

Data are presented as the total number of runs (and the total number of sessions) for analyses on responses to faces/objects.

Transient Inactivation

In the ROI definition process above, for each monkey, we identified four IT face patches (i.e., MF, ML, AF, and AL) using conservative criterion and then focused on these patches as inactivation targets. Reversible inactivation was achieved by infusing muscimol (18 mM, 2.7~3.75 μL, sterile filtered) into each face patch or pair of patches in the awake animals (Turchi et al., 2018). We made microinfusions at the rate of 0.18 μL/min (e.g., for the dosage of 2.7 μL, the injection takes 15 min to complete). After the injection, we waited for 12–15 min before retracting the cannula to avoid any spread of muscimol along the track of the cannula. For additional details on transient inactivation and delimitation of the spread of muscimol, please see our previous study (Liu et al., 2022). Note that the control condition entailed the infusion of saline but all other aspects of the data acquisition remained the same as the muscimol condition.

In our previous study, we showed that the muscimol injection only affected regions in the ipsilateral but not in the contralateral hemisphere, reflecting unilateral connectivity of the face network. It is possible, however, that input from face patch(es) to parietal and frontal regions are propagated to both hemispheres, given the crucial role of these regions in attention, and so we explored the impact of the inactivation on both the ipsilateral and contralateral FEF and LIP.

Data Analysis

Functional data were preprocessed using Analysis of Functional NeuroImages software (20.2.10; Cox, 1996). Images were realigned to the base volume of one initial localizer session. Then, the data were smoothed with a 2-mm FWHM Gaussian kernel. Signal intensity was normalized to the mean signal value within each run. For each voxel, we performed a single univariate linear model fit to estimate the response amplitude for each condition. The model included a hemodynamic response predictor for each category and regressors of no interest (baseline, movement parameters from realignment corrections, and signal drifts). A general linear model and a MION kernel were used to model the hemodynamic response function (Leite et al., 2002).

To investigate the differences in treatment and hemisphere, we conducted generalized linear mixed models (GLMMs) on the data from the single or paired inactivation of middle face patches and anterior face patches separately using SPSS (v24) software (SPSS Inc.). Treatment (control, fundus (F) inactivation, lateral (L) inactivation, and combined F and L inactivation for anterior or posterior patches) was considered a fixed factor, whereas Monkey (C and D), L–R hemisphere (left and right), session, and run were treated as random factors. Subsequently, we carried out post hoc tests on responses to faces/objects in each attention-related ROI, with adjustment for multiple comparisons using the Holm–Bonferroni method. All p values were corrected unless otherwise specified.

RESULTS

We administered muscimol injections, individually or in pairs, into four distinct IT face patches, MF, ML, AF, and AL, spanning the posterior–anterior and fundal-lateral extent of the temporal lobe in two monkeys (Monkey C and Monkey D). Subsequently, using fMRI, we examined the impact of the inactivation of each of the face patches (or pairs), compared with a control condition (saline infusion in the same patch[es]), on two areas related to overt and/or covert visual attention, that is, FEF and LIP. Figure 1B shows the localization of FEF and LIP as well as the face patches in one of the two monkeys (Monkey C). For further details on the number of runs and sessions for the different combinations of face patch inactivations, see Table 1.

We predicted that if there is a causal relationship between face-selective and attention-related areas underlying rapid attentional capture, then the response to faces, but not (or less so) to objects, should be reduced in FEF and/or LIP following middle and/or anterior IT face patch/es inactivation. To evaluate this hypothesis, we conducted a comparative analysis with systematic inactivation of one or a pair of patches, either middle or anterior, versus saline infusion, when animals viewed faces or objects. Below, we present the data from the inactivation of middle face patches, followed by the findings from the inactivation of anterior face patches.

Impact of Inactivation of the Middle Face Patches

To explore the impact of the different inactivation conditions, we performed GLMMs on the fMRI data (normalized beta coefficients) measured from FEF and LIP separately, following the inactivation of middle face patches individually or paired (Figure 2). Condition (control, MF&ML, MF inactivation, and ML inactivation), shown on the x axis, was treated as a fixed factor, and Monkey (C and D), L–R hemisphere (left and right hemisphere), session, and run were considered random factors.

Figure 2.

Effects of middle face patch inactivations on responses in FEF and LIP. The name of the ROI is shown at the top of each graph. (A–B) Effects of middle face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of middle face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of middle face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of middle face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display the mean values + SEM. MF = MF alone inactivation; ML = ML alone inactivation; MF = middle fundal face patch; ML = middle lateral face patch. *p < .05, ***p < .001.

In the hemisphere ipsilateral to the injection site, we found a significant main effect of condition on responses to faces (p < .001) in FEF (Table 2). Post hoc tests revealed that, compared with data from the control sessions, muscimol infusions targeting ML significantly (p < .001) reduced fMRI responses to faces in FEF (Figure 2A). No significant changes were observed in FEF after MF inactivation alone. Notably, we did not find similar changes in FEF following the combined inactivation of MF and ML to the signal reduction after the inactivation of only ML. Note that these data were collected during the later stages of the experiments. As a result of the accumulation of MION, the signals obtained were considerably weaker compared with the data collected earlier in the experiments. Consequently, the variability observed in the weaker signals could potentially account for these inconsistent findings. Increasing the number of runs can mitigate the impact of this variability on the final results. Note that the number of runs during the combined inactivation of MF and ML (n = 32) was much smaller than that for ML inactivation alone (n = 82). Therefore, the reliability of the MF and ML results might be limited.

Table 2.

GLMM Results of the Main Effect of Treatment following Inactivations of Middle and Anterior Face Patches

	df2	Responses to Faces				Responses to Objects
		Ipsilateral Hemisphere		Contralateral Hemisphere		Ipsilateral Hemisphere		Contralateral Hemisphere
		F	p	F	p	F	p	F	p
Inactivation of Middle Face Patches
FEF	298	5.534	<.001	0.317	.813	1.814	.145	1.464	.225
LIP	291	3.423	.018	1.065	.364	2.221	.086	1.900	.130
Inactivation of Anterior Face Patches
FEF	120	9.101	<.001	1.413	.242	13.646	<.001	0.791	.501
LIP	166	3.397	.019	1.493	.218	2.054	.108	1.898	.132

The name of the ROI is shown in the leftmost column.

In LIP, the combined inactivation of MF and ML significantly diminished responses to faces compared with the control condition (p = .032; Table 2 and Figure 2B). Moreover, there was a decreasing trend in face responses in LIP following ML inactivation alone (uncorrected p = .028), whereas MF inactivation alone did not yield a similar effect (uncorrected p = .518). These findings suggest that the impact of the combined inactivation of MF and ML on LIP may primarily stem from the inactivation of ML rather than MF.

The reduction in beta coefficients was restricted to responses to faces: There was no significant effect of condition on responses to nonface objects in either FEF or LIP following the inactivation of middle face patch(es) (Table 2; Figure 2C and D). Furthermore, there were no significant changes as a result of any activation condition on responses to faces or objects in the contralateral hemisphere (Figure 2E–H). It is worth noting that the results from each of the two monkeys are reasonably consistent (see symbols for each monkey in Figure 2 and Figures A3 and 4), with few exceptions. In addition, when the face-selective voxels were removed, similar findings from FEF and LIP were observed (Figure A1).

Impact of Inactivation of the Anterior Face Patches

Next, we investigated the effect of inactivation of anterior face patch(es) on responses to faces and objects in areas FEF and LIP. The same GLMMs were conducted, as described above. In contrast with the effects of inactivating the middle face patch(es), when anterior face patches were inactivated, significant main effects of condition were evident both in responses to faces and to nonface objects in FEF in the hemisphere ipsilateral to the inactivation (Table 2). On post hoc tests, the inactivation of AF alone reduced responses to both faces (p < .001) and nonface objects (p < .001) in FEF. In addition, the combined AF&AL inactivation reduced responses to both faces (p = .035) and to nonface objects (p < .001) in area FEF relative to the control condition (Figure 3A and C). There was no attenuation of responses to faces or objects in FEF caused by the inactivation of AL, indicating that the effects of the combined AF&AL inactivation were likely a result of the inactivation of AF but not of AL.

Figure 3.

Effects of anterior face patch inactivations on responses in the ipsilateral FEF and LIP. The name of the ROI is shown at the top of each graph. (A–B) Effects of anterior face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of anterior face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of anterior face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of anterior face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display mean values + SEM. AF = AF alone inactivation; AL = AL alone inactivation; AF = anterior fundal face patch; AL = anterior lateral face patch. *p < .05, **p < .01, ***p < .001.

In LIP, significant main effects of condition were found in responses to faces (p = .019, Figure 3B) but not objects (p = .108, Figure 3D) in the hemisphere ipsilateral to the inactivation (Table 2). Compared with the control condition, the combined AF&AL inactivation significantly reduced responses to faces (p = .028). There was also a noticeable decrease in responses to faces in LIP following the AL inactivation alone (uncorrected p = .037) but not AF inactivation alone (uncorrected p = .203). This finding suggested that the impact of the combined inactivation of AF and AL on LIP might be attributable to the inactivation of AL rather than AF.

As in the inactivation of middle face patches, compared with the control condition, there were no significant changes in responses to faces and objects in the hemisphere contralateral to the injection site(s) in FEF (Figure 3E–H). As was the case with the findings from the middle face patches, the data from each monkey were largely consistent (see symbols for each monkey in Figure 2 and Figure A5 and A6). Again, similar findings from FEF and LIP were observed when the face-selective voxels were removed (Figure A2).

DISCUSSION

The primary goal of this investigation was to uncover a potential neural correlate for the behavioral phenomenon in which faces, but not other visual stimuli, are subject to rapid and powerful “attentional capture,” either overt or covert. To accomplish this goal, we inactivated four independently identified temporal regions (individually or in pairs) that have stronger neural responses to faces than objects (“face patches”) in NHPs. Of the four patches in the IT cortex, two were situated in the middle, and two were more anterior, with one middle and one anterior patch located in the fundus and the other situated on the lateral surface of the IT cortex (Figure 1B, encircled by black lines). Following the inactivation, we measured the neural response in two key attention-related areas, FEF and LIP (Figure 1B, encircled by green lines), while two monkeys viewed a sequential stream of faces or nonface objects during fMRI scans.

Several major findings emerged, and below, we consider each in turn. These are all schematically depicted in Figure 4. We also include the findings from our previous study (Liu et al., 2022) in Figure 4, in which we illustrate stronger (straight lines) and weaker (dashed lines) causal connections among the four face patches in responses to faces and objects. We have now superimposed the findings from the current investigation onto this circuit and showed the consequences of inactivating middle and anterior face patches on FEF and LIP.

Figure 4.

Summary of the connections between the face patches and areas FEF and LIP in the ipsilateral hemisphere. Data are shown here on the cortical surface of the right hemisphere. There were no significant effects in the hemisphere contralateral to the injection. The red and blue lines with arrows are imported from our previous article about the internal connections of the face patches: red lines = effects on responses to faces; blue lines = effects on responses to objects; red dashed lines = weak or variable effects on responses to faces; blue dashed lines = weak or variable effects on responses to objects; black dashed lines = weak or variable effects on responses to faces and objects. The orange and cyan lines with arrows are from the current study and indicate the impact of the inactivation of IT face patches on responses in attention-related regions: orange lines = effects on responses to faces; cyan lines = effects on responses to objects; orange dash lines = weak or variable effects on responses to faces. MF = middle fundal face patch; ML = middle lateral face patch; AF = anterior fundal face patch; AL = anterior lateral face patch.

In a nutshell, the findings indicate that the circuit that has the highest likelihood of mediating attentional face capture is between the middle lateral face patch, ML, and the attention-related regions, FEF, and, to a lesser extent, LIP, specifically when faces are viewed (Figure 2). Inactivation of the anterior fundal face patch, AF, and inactivation of AF in combination with the anterior lateral patch, AL, reduced responses to faces as well as objects in FEF, whereas inactivation of AL, and in combination with AF, affected LIP responses to faces but not objects. In addition, the inactivation of a single face patch or a pair of face patches appears to have minimal to no discernible consequences on responses of FEF or LIP in the contralateral hemisphere. The results from each of the two monkeys were largely consistent in the conditions described here, offering support for the conclusions we have reached.

Together, these empirical findings uncover a potential neural network linking IT face patches and two attention-related regions, which may underly the attentional capture of faces. Additional investigation of this circuit in an explicit attentional face capture task and confirmation of the connectivity and contribution of other regions that might play a mediating role (e.g., frontal face patches) are needed to validate and perhaps extend our findings.

Impact of Inactivation on Ipsilateral versus Contralateral Responses

The first general finding is that the inactivation of a face patch or a pair of face patches has minimal to no consequences on FEF and LIP in the contralateral hemisphere. Instead, the reduction in neural response to viewing images (faces or nonface objects) arises mainly from the inactivation of face patch(es) in the temporal lobe ipsilateral to FEF and LIP.

In our previous study (Liu et al., 2022), we also showed that inactivation of face patches led to attenuation of the response in other face patches on the ipsilateral but not the contralateral side. Consistently, neural activation in monkeys induced by microstimulation is also typically more robust in the hemisphere ipsilateral to the site of stimulation (Moeller et al., 2008). We did not assume that this would necessarily also be true for face-attention circuitry, and so, we included both hemispheres in this investigation. Indeed, one recent study revealed that some regions in monkeys do have symmetric coupling and direct interhemispheric connectivity; for example, as recently reported, the right and left middle lateral patches, but not the anterior face patches, are symmetrically coupled (Zaldivar et al., 2022), further motivating us to consider both hemispheres. Our findings are clear, however: Relative to a saline control infusion, there is minimal to no change to FEF and LIP responses in the contralateral hemisphere following inactivation of any face patch(es). These findings attest to the specificity of the effect of the muscimol infusion and serve as a strong within-monkey control condition.

Differences between Relationships of FEF and LIP with IT Face Patches

The key result concerns the differential effects of IT face patch inactivation on FEF and LIP. First, ML inactivation led to attenuated neural responses to faces in FEF and, to a lesser extent, LIP. Second, the impact of the inactivation of anterior face patches on FEF and LIP was obvious, but there was a difference in profile between the IT face patches and their associations with FEF and LIP: AF might be the primary source of the effects of the combined AF&AL inactivation on FEF, whereas AL might contribute to the impact of the combined AF&AL inactivation on LIP. We consider these connectivity differences further below.

Relationships of FEF with IT Face Patches

Previous tracer studies in monkeys have demonstrated distinct projection patterns from the STS to the FEF: the dorsal bank of the STS projects to the medial FEF, the ventral bank projects to the lateral FEF, and the fundus projects throughout the entire FEF (Saleem, Miller, & Price, 2014; Schall, Morel, King, & Bullier, 1995). Notably, the face patches AF and MF are located within the fundus of the STS. Consistent with the previously identified anatomical connections, we found that FEF likely receives inputs from the face patches, particularly from area AF (Figure 4). Although previous tracer studies using retrograde tracers injected into face patches have reported weak projections from the PFC, tracers were not injected into AF and MF in that study (Grimaldi, Saleem, & Tsao, 2016). Therefore, our current study provides direct evidence supporting the transfer of visual information from the anterior fundal face patches to the PFC.

In addition, the locations of AL and ML are outside the areas reportedly connected with the FEF ROI defined in our study (Saleem et al., 2014; Schall et al., 1995). This leads to the prediction that inactivation of either of these patches should have no downstream effect on FEF. Although the inactivation of ML did impact the responses to faces in FEF (Figure 4), this influence may be mediated through an intermediate station, such as AF. It is worth mentioning that there are existing connections between AF and ML, albeit somewhat weak (Liu et al., 2022; Moeller et al., 2008). A future study adopting the dual inactivation and fMRI approaches might inactivate AF and ML as a pair (and AL and MF) to better establish the nature of the connectivity to FEF and to determine more conclusively whether the connectivity is direct or indirect.

Relationships of LIP with IT Face Patches

In the present study, we found that the inactivation of anterior and middle face patches had an impact on visual responses to faces in LIP, which likely is attributable to the lateral face patch (AL and ML) inactivation. Previous tracer studies have identified connections between LIP and area TEO (especially the lateral area; Webster, Bachevalier, & Ungerleider, 1994; Distler, Boussaoud, Desimone, & Ungerleider, 1993; Blatt, Andersen, & Stoner, 1990), in which the middle face patches are located. Specifically, reciprocal cortico-cortical connections have been identified between the ventral portion of LIP (LIPv) and TEO (Blatt et al., 1990). Other studies have also indicated connections between both the ventral and dorsal portions of LIP (i.e., LIPv and LIPd) with TEO (Webster et al., 1994; Distler et al., 1993). In addition, there are connections between LIP and TE, wherein the anterior face patches are located, although these connections are relatively limited compared with those with TEO. For example, sparse labeling in LIP was found after tracer injections were performed in TE (Morel & Bullier, 1990; Shiwa, 1987). Note that, in most of these studies, tracers were injected into the lateral portions of TEO and TE (Webster et al., 1994). In the present study, our results show that the inactivation of ML and AL might influence responses in LIP. Note that our previous study (Liu et al., 2022) revealed that the face-processing network is organized along a posterior–anterior axis with a bidirectional dialogue.

In the present study, we observed that the inactivation of middle face patches led to a reduction in responses specifically to faces, while leaving object responses relatively unaffected, as hypothesized. However, the inactivation of anterior face patches resulted in a reduction in responses to both faces and objects in FEF. Two possibilities may explain these findings. On the one hand, the impact of inactivating the IT face patches may function through the modulation of face- (or shape-)related neurons in FEF. In macaques, several electrophysiological studies have documented the presence of face- or shape-related activity in the ventrolateral PFC, specifically in the inferior prefrontal convexity (Ó Scalaidhe, Wilson, & Goldman-Rakic, 1997, 1999; Pigarev, Rizzolatti, & Scandolara, 1979), as well as in distinct regions of the parietal cortex, such as LIP and anterior intraparietal area (Lehky & Sereno, 2007; Sakata et al., 1998; Sereno & Maunsell, 1998). In addition, a previous fMRI study has reported the presence of face-selective clusters in FEF (and LIP) (Rajimehr et al., 2009). In particular, in line with previous studies, we successfully identified the frontal face patches, one of which was located adjacent to the arcuate sulcus (prefrontal arcuate face patches [PA]), as depicted in Figure 1B, in close proximity to FEF. Although the FEF and LIP ROIs without voxels exhibiting face-selective responses yielded similar results (Figures A1 and A2), there was no clear boundary between face-selective and non face-selective voxels, which could be attributed to a gradient distribution of face-selective neurons decreasing from the center outward, a phenomenon previously observed in IT face patches (Bell et al., 2011). Hence, the attenuating effects on responses to objects in addition to faces observed in FEF may be attributable to the responses of face-related neurons within it. These neurons could potentially receive information about faces and objects from the anterior face patches. Further investigations are warranted to address this question more comprehensively in future studies.

Alternatively, although the inactivation successfully targeted the face patches, the injected muscimol covered them very well, though not perfectly (see details in Liu et al., 2022). In the IT cortex, face patches are surrounded by nonface selective regions, such as object-selective regions. Thus, although we did not observe any significant influence of IT face patch inactivation on the object-selective regions in IT cortex and LIP, there may still exist an effect, as observed in our study for FEF.

Methodological Considerations

In this study, we used transient inactivations of specific brain regions in combination with fMRI. Other methodological approaches might also be used to examine the connectivity between IT and frontal and parietal areas, for example, neural stimulation (microstimulation) and single-unit recording. Both approaches have their pros and cons. One complicating property of neural stimulation is that its effects are transynaptic and bidirectional, making inferences about the directionality of transmission difficult (Klink, Dagnino, Gariel-Mathis, & Roelfsema, 2017; Moeller et al., 2008). We, therefore, elected to use fMRI because we hoped to gain insight into the nature of signal propagation between IT nodes and other regions. In addition, we chose fMRI because it allows the simultaneous measurement of responses from multiple areas, which is difficult to achieve using electrophysiological recordings in the same animal. In our study, we focused on FEF and LIP, which are located far apart from each other, as well as far from the IT cortex, and we conducted these investigations in both hemispheres. Using electrophysiological recordings for such parallel investigations would require the use of multiple chambers or multielectrode implants. Both of these approaches would increase surgical complexity, and the latter approach can affect the integrity of the ROIs to be examined (Patel et al., 2023). That said, we anticipate that studies using targeted recordings as well as studies using fMRI in NHPs will continue to advance our understanding of neural mechanisms underlying attentional capture of faces.

We also elected to use saline infusion and fMRI during stimulus-evoked responses as a baseline to mimic the experimental inactivation condition as closely as possible, including the local mechanical perturbation. Thus, the only difference is that in the experimental condition, the infusion contained muscimol to induce the inactivation. Many other control baselines might be used in such a study and likely would yield valuable insights, including the use of resting-state fMRI. We rejected the resting state option, given the fMRI signal as well as connectivity might be weaker in resting state than in task-evoked conditions (Avidan & Behrmann, 2009, 2021).

Finally, we included two participants in our study. Recording from two monkeys is in keeping with many, perhaps even the majority of, investigations with NHPs to minimize the number of animals, in line with the 3Rs (replacement, reduction, and refinement) requirements (https://www.ncbi.nlm.nih.gov/books/NBK54050/). The number of participants has been hotly debated for 2 decades, with earlier studies calling for a larger number of participants (Murphy & Garavan, 2004), whereas more recent studies emphasize reliability at the single-subject level (Stevens, D’Arcy, Stroink, Clarke, & Beyea, 2013; Savoy, 2006). As Savoy (2006) noted, “It may someday turn out that the information from a few brains, thoroughly studied, will reveal more about universal aspects of human brain function and organization than the current torrent of studies from large collections of brains.” We elected to collect data from two monkeys under a wide range of manipulations and carefully controlled experimental conditions. Unsurprisingly, although in some cells of the experiment, the data were sparse, but collectively, the possibility of “deep phenotyping” and obtaining comprehensive data trumped the advantages afforded by a multisubject study. We also note that previous studies employing similar techniques have likewise limited their subject enrollment to only two individuals (Bogadhi, Bollimunta, Leopold, & Krauzlis, 2019; Miyamoto et al., 2017).

Limitations and Future Directions

We have characterized a widely distributed face network in two monkeys and have measured the attenuation of attentional areas following the inactivation of parts of this widespread circuit in a joint perturbation-measurement design. In some inactivation conditions (e.g., the combined MF&ML inactivation), given the more limited data as a function of MION accumulation, we had to concede that the results were not fully conclusive. Nevertheless, we have suggested a potential candidate circuit that may give rise to the attentional capture of faces in monkeys and humans.

A further limitation is that we did not test the attentional capture of faces directly and did not probe the underlying neural mechanism in parallel with behavior as would be required to reach any clear conclusions. Rather, we opted for an approach that would enable us to reveal the necessary nodes of a circuit that might be implicated in attentional capture. Our results have, in fact, allowed us to propose such a circuit. The coherent and strong prediction is that in both humans and NHPs, attentional capture of faces may engage the AF-FEF and ML-LIP circuits. Of course, as we have stressed repeatedly thus far, a more detailed evaluation of the attention-face patch circuit is required under conditions that specifically engage the attentional capture of faces.

We also were not able to examine the frontal face patches that have been identified in both Old World (Tsao, Schweers, et al., 2008) and New World monkeys (Schaeffer et al., 2020). Although we were able to successfully identify the frontal face patches, such as PA, during the initial localizer task, the subsequent experiments presented challenges in consistently identifying these face patches because of MION accumulation. Consideration of the contribution of the frontal face patches would further delimit the face patch-attention circuit and help refine the specific circuitry (and potential face patches that might serve as intermediaries) that we have suggested based on the current investigation. Finally, as alluded to previously, information about social arrangements, including face familiarity and positive relationships (Lewis et al., 2023) might well modulate attentional capture too, and is deserving of further investigation to unravel the full complexity of the circuit described herein.

Conclusion

The results we have presented offer a potential face-attention circuit that might underlie the attentional capture of faces, at least as uncovered in NHPs. Comparable studies using fMRI have not been done in humans, and no perturbation studies have been conducted. Given the widespread agreement regarding homologies between humans and NHPs (Yovel & Freiwald, 2013; Yovel & Kanwisher, 2004), we tentatively suggest that the findings of the present study may well have application for understanding the neural circuit engaged in the human attentional capture of faces. Both human and NHP species have multiple face patches and also have hierarchical and parallel pathways for face processing, consistent with the claims of homology. A future perturbation-type study with humans using TMS on fMRI-identified face patches (although access to deeper and inferior structures is difficult) and/or a study of individual(s) with a focal, circumscribed lesion affecting just one IT face patch, might shed further light on cross-species homologies. Similar methodologies might be used to investigate attentional effects. Finally, fMRI data acquired during an attentional face capture paradigm will contribute further to our understanding of the face-attention circuit that is the focus of the current investigation.

APPENDIX

Figure A1.

Effects of middle face patch inactivations on responses in FEF and LIP without face-selective voxels. The name of the ROI is shown at the top of each graph. (A–B) Effects of middle face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of middle face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of middle face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of middle face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display the mean values + SEM. MF = MF alone inactivation; ML = ML alone inactivation; MF = middle fundal face patch; ML = middle lateral face patch. *p < .05, ***p < .001.

Figure A2.

Effects of anterior face patch inactivations on responses in the ipsilateral FEF and LIP without face-selective voxels. The name of the ROI is shown at the top of each graph. (A–B) Effects of anterior face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of anterior face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of anterior face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G-H) Effects of anterior face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display mean values + SEM. AF = AF alone inactivation; AL = AL alone inactivation; AF = anterior fundal face patch; AL = anterior lateral face patch. *p < .05, **p < .01, ***p < .001.

Figure A3.

Effects of middle face patch inactivations on responses in FEF and LIP without face-selective voxels in Monkey C. The name of the ROI is shown at the top of each graph. (A–B) Effects of middle face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of middle face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of middle face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of middle face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display the mean values + SEM. MF = MF alone inactivation; ML = ML alone inactivation; MF = middle fundal face patch; ML = middle lateral face patch.

Figure A4.

Effects of middle face patch inactivations on responses in FEF and LIP without face-selective voxels in Monkey D. The name of the ROI is shown at the top of each column. (A–B) Effects of middle face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of middle face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of middle face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of middle face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display the mean values + SEM. MF = MF alone inactivation; ML = ML alone inactivation; MF = middle fundal face patch; ML = middle lateral face patch. *p < .05, **p < .01, ***p < .001.

Figure A5.

Effects of anterior face patch inactivations on responses in the ipsilateral FEF and LIP without face-selective voxels in Monkey C. The name of the ROI is shown at the top of each column. (A–B) Effects of anterior face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of anterior face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of anterior face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of anterior face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display mean values + SEM. AF = AF alone inactivation; AL = AL alone inactivation; AF = anterior fundal face patch; AL = anterior lateral face patch. *p < .05, **p < .01, ***p < .001.

Figure A6.

Effects of anterior face patch inactivations on responses in the ipsilateral FEF and LIP without face-selective voxels in Monkey D. The name of the ROI is shown at the top of each column. (A–B) Effects of anterior face patch inactivation on responses to faces in (A) FEF and (B) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (C–D) Effects of anterior face patch inactivation on responses to nonface objects in (C) FEF and (D) LIP in the hemisphere ipsilateral to the inactivation sites across both monkeys. (E–F) Effects of anterior face patch inactivation on responses to faces in (E) FEF and (F) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. (G–H) Effects of anterior face patch inactivation on responses to nonface objects in (G) FEF and (H) LIP in the hemisphere contralateral to the inactivation sites across both monkeys. Bars display mean values + SEM. AF = AF alone inactivation; AL = AL alone inactivation; AF = anterior fundal face patch; AL = anterior lateral face patch field. *p < .05, **p < .01, ***p < .001.

Diversity in Citation Practices

Retrospective analysis of the citations in every article published in this journal from 2010 to 2021 reveals a persistent pattern of gender imbalance: Although the proportions of authorship teams (categorized by estimated gender identification of first author/last author) publishing in the Journal of Cognitive Neuroscience (JoCN) during this period were M(an)/M = .407, W(oman)/ M = .32, M/W = .115, and W/ W = .159, the comparable proportions for the articles that these authorship teams cited were M/M = .549, W/M = .257, M/W = .109,andW/W= .085 (Postle and Fulvio, JoCN, 34:1, pp. 1–3). Consequently, JoCN encourages all authors to consider gender balance explicitly when selecting which articles to cite and gives them the opportunity to report their article’s gender citation balance.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, upon reasonable request.