Attention represents an important avenue of exploration in Autism Spectrum Disorder (ASD) given that individuals have consistently been observed to have impairments [1]. In one of the largest studies undertaken, up to 93% of observed participants were shown to have some form of attentional deficit [2]. While a diagnosis of ASD was previously grounds for exclusion from receiving a diagnosis of attention deficit hyperactivity disorder (ADHD) (DSM IV), the growing awareness of the overlap in attention deficits between the two disorders [1, 3, 4] has caused this to change (DSM V). Deficits in attention for ASD are commonly observed in social contexts (e.g., failures of joint attention and turn taking) and are less severe for executive functioning than in ADHD [3]. Attentional deficits in ASD, however, also occur in elementary forms of processing, including simple target detection tasks [5] and tests of vigilance or sustained attention [6]. For more information on the commonalities and differences in attentional deficits between ASD and ADHD, see [7].

The concept of attention is, in itself, a series of operations that integrate external sensory inputs with internal information and decision-making and many subcomponents are required to complete even the most basic task [8, 9]. Now that attentional deficits have been demonstrated to be a consistent finding in ASD, it is therefore worthwhile to isolate the subcomponents of attention in an attempt to determine if these impairments represent a wholesale dysfunction or localized deficits. In doing so, we may be able to better tailor our management styles to aid those affected. To this end, the attention model of [8] has served as a basis for identifying the subcomponents of attention in previous investigations [10-14].

This model suggests that there are three main subcomponents of attention: alerting, orienting, and executive control [8]. Each subcomponent is thought to represent an independent neural circuit and is composed of different regions of interest (ROIs) interacting with one another [8]. In brief, the locus coeruleus, right frontal and right parietal cortices are associated with alerting, the temporoparietal junction, superior parietal cortex, and frontal eye fields are associated with orienting, and the prefrontal cortex and anterior cingulate cortex (ACC) are associated with executive control [15].

In order to assess the three subcomponents of attention proposed by Posner, et al. [8] (i.e., alerting, orienting, and executive control), Fan et al. [10] created the Attentional Networks Task (ANT). The task consisted of a series of arrows pointing either right or left with a “target” arrow at the center of the screen. Participants were asked to identify the direction of the centrally presented arrow by pressing a key indicating either the left or right direction. For the alerting condition, an asterisk was briefly flashed at the center of the screen before the stimuli were shown, thus alerting participants to an upcoming stimulus. For the orienting condition, two warning cues corresponding to the position of the upcoming stimuli were briefly flashed at either the top or bottom of the screen, thus orienting participants to the position of the upcoming stimuli. Finally, executive control was evaluated by having flanking arrows either be congruent or incongruent with the direction of the target (i.e., central) arrow [10]. The ANT was originally tested in neurotypical (NT) adults and was shown to produce consistent single-subject estimates of the three subcomponents of attention. These results further supported the notion that processing efficiencies for the three subcomponents is largely independent [10].

A variation of this test designed for children (aptly named the Child Attention Network Task; Child ANT) was then created for the possibility of providing insight into how to help improve their attentional abilities [16]. The Child ANT has since been used to investigate attention processes in NT children [17], children with attention deficit hyperactivity disorder (ADHD) [12], and children with ASD [14].

Previous investigations of ASD through the use of the ANT and Child ANT have shown differing results with deficits being observed in all three subcomponents of attention at varying ages [11, 13, 14]. Posner and Rothbart [15] suggested that the orienting network deficiencies (perhaps the most consistent finding in ANT studies of ASD) are a probable explanation for the commonly reported overly fixative nature of those with autism (e.g., their tendency to focus their gaze on an ear lobe rather than on more informative features of a human face such as the eyes). Others, however, have proposed that these delays are in reality due to executive impairments [18, 19], which produce difficulty in dividing (or inhibiting) attentional resources amongst competing external stimuli. Neuroimaging may help settle this dispute by allowing us to observe the areas of activation whilst ASD participants perform the ANT in real time.

Indeed, fMRI studies of ASD have exploded in recent years and range in topics from social cognition to inhibition control [20]. Amongst the most consistent findings related to executive control are atypical activations in the ACC in participants with ASD [21-23], suggesting decreased cortical specialization (mild shifting of cortical location areas in response to a variety of tasks), leading to the hypothesis that autism is a distributed brain systems disorder [24]. (An extensive review can be found in [24]). A previous fMRI investigation of attentional processing in children with ASD used socially and non-socially-cued stimuli to show that regions commonly activated in NT children for the socially-cued stimuli were only active for children with ASD during the non-socially-cued stimuli presentations. Unique areas of activation in the left dorsal frontal parietal cortex and medial temporal lobe were seen when children with ASD saw the socially-cued stimuli ([25]. Therefore, the incorporation of the ANT with fMRI may allow us to identify differences in attentional processing between adolescents with ASD and their NT counterparts for the subcomponents of attention.

The present study uses the child ANT [16] in combination with functional magnetic resonance imaging (fMRI) to identify behavioral and brain activation differences between ASD and NT adolescents for the three subcomponents of attention in an attempt to reconcile previous observations and to determine if the consistently seen deficits of orienting attention are truly due to dysfunctions of orienting-based networks or are a reflection of executive dysfunctions. In accordance with previous work in children, we expected that ASD adolescents would perform demonstrably worse on the orienting and executive portions of the Child ANT, while performance on the alerting section would likely be normal [14]. We hypothesized that deficits in orienting and executive performance would relate to decreased patterns of brain activity within the fronto-cerebellar network (specifically the ACC, the middle and superior frontal gyri (MFG and SFG), and the cerebellum); a network previously indicated to play a role in directing attention (i.e., searching for responses) and executive decision-making (i.e., response selection) [26] more so than the orienting network proposed by [8]. We base this assertion on consistent pathology and decreased brain activity noted in the cerebellum of ASD individuals as summarized in a review by [27].

Six high-functioning ASD (Mean age: 17.33 years, 4 males/ 2 females, ADOS scores ranging from 7 to 21) and six NT adolescents (Mean age: 15.83 years, 4 males/ 2 females) performed the Child ANT while undergoing fMRI scanning. Clinical diagnoses were confirmed using the Autism Diagnostic Observation Schedule (ADOS). The ASD group did not display significant motion artifacts for any participants and standard preprocessing motion correction using MCFLIRT was performed on all subjects. The response rate was 100% for both groups. Per parent-report, participants in the NT group were free of ASD-related symptoms or any other neurological or psychiatric conditions. Participants provided informed consent and received $25 compensation for their involvement in the study. All participants had normal or corrected-to-normal vision and were right-handed. Informed written consent was obtained from all participants before scanning. This study was approved by the Institutional Review Board (IRB) of Texas Tech University.

Six high-functioning ASD and six NT adolescents performed the Child ANT while fMRI images were acquired via a Siemens 3T Skyra. A modified version of the Child ANT was developed for use inside of the scanner and was presented using ePrime 2.0. Each session lasted 30 minutes and consisted of 3 blocks of 48 trials. Stimuli were presented for 2 second-intervals in an event-related design. Participants were told to “feed the fish” by pressing fiber optic buttons (left or right index finger) that matched the direction the target fish was facing. Trials began with a fixation cross followed by 1 of 4 trial types, each indexing the efficiency of alerting, orienting and executive attention by presenting stimuli in 4 different cueing conditions: no cue, central cue, double cue, and spatial cue. On no cue trials, the target fish appeared alone; central cue, an asterisk was presented at the same location as the fixation cross; double cue, an asterisk appeared simultaneously above/below the fixation cross; spatial cue, a single asterisk was presented in the same position as the target fish. On executive trials two flanker fish appeared to the left/right of the target, oriented either congruently/incongruently (i.e., pointing in the same/opposite direction, Fig. 1).

Fig. (1). Schematic of the experimental protocol for the Child ANT (adapted from Rueda et al. 2004) with the target stimulus being the center yellow fish.

Response time (RT) and accuracy were calculated for each attentional subcomponent using the methodology developed by [10] alerting (no cue - center cue), orienting (center cue - spatial cue) and executive (congruent flanker - incongruent flanker). A 2(group) x 3(attention type) mixed-design analysis of variance (ANOVA) was performed for the RT and accuracy data.

Functional images were acquired via a Siemens 3T Skyra (located at the Texas Tech Neuroimaging Institute; TR=2000ms; TE=20; FOV=240 mm; Flip angle=80; Voxel size: 3x3x4 mm; axial slices=34). Structural 3D high resolution T1 images were acquired using same orientation as the functional sequences (0.9X0.9X0.9 mm). Functional data analyses were carried out using the fMRI Expert Analysis Tool (FEAT) in FSL. Imaging preprocessing involved removing non-brain structures by Brain Extraction Tool (BET), motion correction by using MCFLIRT, temporally high-pass filtered with a cutoff period of 100s, spatial smoothing with a 5mm Gaussian FWHM algorithm and co-registration of the fMRI data to anatomic images with each subject and MNI 152 standard brain space.

The Child ANT paradigm was presented using ePrime with a random ordering of trials per session. The log files for each session were interrogated to determine the order of trials and an onset time vector was generated for each of the six conditions in that session. The design matrix was constructed to model the three attention conditions at the individual level, by segregating images according to attention/trial type (i.e., alerting, orienting, and executive). These vectors were grouped according to the above three network definitions and saved along with duration values and condition labels for importation during the First Level Analysis (FLA). Since this study was event-driven, durations were determined from the log files, as they were dependent on participant responses. A summary of the fMRI contrasts used to determine each subcomponent network is given in Table 1.

In the group level analyses, average brain activations across participants and contrasts between groups were conducted using the FMRIB Local Analysis of Fixed Effects tool (FLAFE) and activation was thresholded using clusters determined by z >2.3 at P_corrected<0.05 (default FWE correction in FSL).

A 2(group) x 3(attention type) mixed-design analysis of variance (ANOVA) on the accuracy data revealed no significant main effects, but did reveal a significant Group x Attention Type interaction (F= (2, 20) =6.307, p < 0.01). A 2 (group) x 3 (attention type) mixed-design ANOVA on the RT data revealed no significant main effects and no significant interactions with attention type (Fig. 2a). Post-hoc comparisons of least significant difference test (LSD) (p < 0.05) showed a significant difference on alerting and orienting accuracy between the ASD and NT groups (Fig. 2b). Specifically, ASD participants (M=-1.852, 95% CI [-3.609,-0.095]) were more accurate than NT participants (M=1.389, 95% CI [-0.368, 3.146]) on alerting trials, while NT participants (M=-1.389, 95% CI [-2.610, -0.168]) were more accurate than ASD participants (M=0.463, 95% CI [-0.758, 1.684]) on orienting trials. There was no significant difference between the groups on executive trials. NT and ASD adolescents showed no difference in RT for any of the three attentional networks.

Fig. (2). Behavioral results of the CHILD ANT tasks. (a) RT difference scores for ASD and NT groups for each attentional type: alerting (blue), orienting (green), and executive control (beige). Note: the higher values indicate more efficient performance. (b) Differences in accuracy scores for ASD and NT groups for each attentional type: alerting (blue), orienting (green), and executive control (beige). Note: the lower values indicate more efficient performance. Indicates Post-hoc comparisons of least significant difference test (LSD) (p<0.05)

(Fig. 3) presents the activation areas of significant contrast for the ASD versus NT adolescents. In the alerting attention condition, more activation was found for the NT group in the left cerebellum (i.e. the left inferior semi-lunar lobule, the left pyramis, the left culmen, and the left declive) and in the left parahippocampal gyrus (Brodmann Areas (BA) 19 and 36). No significant differences in alerting activation were found in the ASD minus NT contrast. For orienting attention, the ASD group was significantly more active than the NT group in the bilateral MFG (BA 9 and BA 10), the right SFG (BA 10) and the left ACC (BA 32). Again, no significant activation differences were found between groups for the ASD minus NT contrast. For executive attention, the ASD group was significantly more active than the NT group bilaterally in the anterior cerebellum. The NT minus ASD contrast revealed no significant differences in activation between the two groups. Table 2 lists the exact coordinates of activation for each of the three attention types as a function of ASD versus NT contrasts.

Fig. (3). Activation maps for contrasts of group averaged BOLD fMRI signal intensities for ASD and NT groups. Axial slices demonstrate average brain activations across participants and contrasts between groups using the FMRIB Local Analysis of Fixed Effects tool (FLAFE) and thresholding clusters of z >2.3 at Pcorrected<0.05 for: alerting (top), orienting (middle), and executive control (bottom).