Searching for the attention deficit in attention deficit hyperactivity disorder: The case of visuospatial orienting

Abstract

We review all 14 extant studies of covert visuospatial attention in attention deficit hyperactivity disorder (ADHD) (total N=248). Metaanalysis showed that intriguing but isolated findings of alerting or posterior disengage deficits were too small to reliably detect with the sample sizes typically employed. Posterior move and engage operations and the vigilance sustained attention process were normal in ADHD. For exogenous cues, effect sizes for group differences were homogeneously small across all repeated-measures conditions, as were calculations of cost, benefit, and validity effects. For endogenous cues, effect sizes were heterogeneous; however, calculations of cost, benefit, and validity effects were small and homogenous. The most parsimonious conclusion may be that ADHD is not characterized by significant visual orienting dysfunction, but questions remain about the extent of anterior lateralized effects in the combined subtype and about attentional functioning in the inattentive subtype.

Introduction

As its name implies, attention deficit hyperactivity disorder (ADHD) has been thought for at least two decades to involve a possible deficit in attention. The concept of attention within the diagnostic criteria of DSM-IV (American Psychiatric Association, 1994) is not formally defined in cognitive terms. Instead, behavioral ratings from parents and teachers reliably differentiate two symptom domains labeled as inattention-disorganization and hyperactivity-impulsivity (McBurnett et al., 1999). However, behavior that appears as “inattentive,” spacey, off task, and the like, may or may not be traceable to dysfunction in basic cognitive or neural networks that subserve attentional control per se. As others and we have pointed out, such behaviors could be related to a variety of candidate cognitive dysfunctions (Nigg, 2001). Controversy remains in the literature as to whether cognitively defined attentional processes are in fact dysfunctional in ADHD (Barkley, 1997b).

Complicating resolutions of such questions are the changing phenotypic descriptions and diagnostic criteria for ADHD. DSM-III (American Psychiatric Association, 1980) included subtypes of “ADD” with and without hyperactivity. DSM-III-R (American Psychiatric Association, 1987) removed this distinction and described “ADHD” as a single disorder of inattention, hyperactivity, and impulsivity (Faraone, Biederman, Weber, & Russell, 1998). Following multiple studies and extensive field trial investigations that supported a two-factor structure to ADHD symptoms (inattention-disorganization vs. hyperactivity-impulsivity) and clinically meaningful differences between subtypes, DSM-IV (American Psychiatric Association, 1994) returned to subtyping Faraone et al., 1998, Milich et al., 2001. However, most of the studies reviewed herein relied on DSM-III-R criteria. Unless otherwise specified, when we refer to “ADHD,” we are referring to the disorder in general without regard to subtype.

It may surprise some specialists that we claim the status of attention in ADHD has not already been settled in the negative, while clinicians may be surprised that attention is not an established deficit. Many current theories no longer emphasize attention processes, instead highlighting deficits in executive functions (EF) (Barkley, 1997b), arousal (Zentall & Zentall, 1983), allocation of effort during motor output (Sergeant, Oosterlaan, & van der Meere, 1999), reward response (Newman & Wallace, 1993), or other disturbance (for a review, see Nigg, 2001). Yet, theorists remain concerned about attentional dysfunction in ADHD (e.g., Douglas, 1999), and attentional deficits continue to receive extensive empirical investigation, suggesting that scientists and clinicians remain curious as to the relationship between inattention and ADHD.

The continued interest in attention may be due to several factors including the promise of future assessment tools that can better characterize a core deficit marking the disorder (Neufeld, 2002). More generally, it reflects the movement in the scientific field toward a search for cognitive markers or other endophenotypes that may mark etiologically significant dysfunctions. Advances in cognitive neuroscience, and particularly the development of a model emphasizing component operations involved in visuospatial orienting, have revived the study of attentional processes in ADHD. This model, in turn, derives from seminal work by Posner (1980) and others. After the pioneering study of this model in ADHD by Swanson et al. (1991), 13 studies followed using related although not identical methods. As these efforts have gone forward, it is timely that the status of attentional orienting in ADHD be reviewed.

As context for this review, we note that several other approaches to attention are apparent to even a casual observer of the cognitive neuroscience (and even to some degree the clinical) literature in the past decade. These include new advances in the analysis of load-dependent perceptual processing Huang-Pollock et al., 2002, Lavie, 1995, conflict detection and resolution of interference (Botvinick, Braver, Barch, Carter, & Cohen, 2001), and working memory (Awh & Jonides, 2001) among others. It would require a book-length review to scrutinize potential applications of all of these approaches to ADHD or other psychopathology. However, in the last 10 years, arguably the most heavily relied on approach to attention in studies of ADHD has been based on the covert spatial orienting model using a particular orienting measurement paradigm.

We evaluate whether a deficit in any of the distributed neural systems described in that model is likely to characterize ADHD using that paradigm. In planning this review, we noted the clear benefits of quantitative metaanalysis for estimating population effects. We also noted that the literature we review is exceptionally diverse in its methodology, requiring some qualitative evaluation as well. We therefore decided to pursue both approaches herein. Because it forms the basis for all studies reviewed, we first describe the model of attentional orienting.

As outlined by Posner and Petersen (1990), visual attentional control comprises at least three distributed neural systems: the anterior attention system (AAS), the posterior attention system (PAS), and the vigilance system. The AAS, or executive attention network (Posner & Raichle, 1994), is a supervisory system neuroanatomically including the anterior cingulate gyrus, supplementary motor cortex, and other areas of the midprefrontal cortex Jackson et al., 1994, Mirsky, 1996. It is viewed as responsible for exercising deliberate control over information processing (Posner & Raichle, 1994), including the voluntary shifting of attention to locations in space Jackson et al., 1994, Jonides, 1981. These functions can be thought of as similar to various other conceptions of EFs in the ADHD neuropsychological literature Barkley, 1997a, Barkley, 1997b, Nigg, 2000, Nigg, 2001, Pennington & Ozonoff, 1996. It is consistent with theories of global executive dysfunction to expect ADHD deficits in the AAS. We refer to the AAS as supporting voluntary orienting of visuospatial attention.

Depending on context, visual attention is also automatically directed. A second distributed network, termed the PAS, subserves this function. Neuroanatomically, the PAS receives extensive norepinephrine-rich projections from the locus coeruleus and involves the superior parietal cortex, pulvinar, and superior colliculus Posner & Raichle, 1994, Rothbart et al., 1994. Through norepinephrine inputs, the PAS serves to automatically orient attention to sudden changes in the perceptual field (e.g., a bright light or moving object). Studies of patients with focal lesions show distinct regions involved in automatically engaging, disengaging, and shifting spatial attention Posner & Raichle, 1994, Posner et al., 1987, Rafal & Posner, 1987.

The third system in this model is the vigilance system. Vigilance is often referred to as sustained attention, which has its own extensive history of investigation (Mirksy & Duncan, 2001). It is defined as the positive ability to maintain a tonic state of alertness and wakefulness during prolonged and sustained mental activity (Weinberg & Harper, 1993). Performance that deteriorates over the course of minutes or hours (operationalized as slower reaction times [RT], greater variability of response, and increased errors) is known as the “vigilance decrement” (Parasuraman, Warm, & See, 1998). A second function of this system can also be discerned. A deficit in “alerting” or phasic arousal occurs with poor performance (slow, variable response times) either on a series of single abrupt onset trials or poor performance from the very beginning of a time course. Posner et al. have argued that phasic arousal and vigilance are both part of the “vigilance system” and that both depend on the same network of neural structures including the noradrenergic system of the locus coeruleus, the cholinergic system of the basal forebrain, the intralaminar thalamic nuclei, and the right prefrontal cortex Parasuraman et al., 1998, Rothbart et al., 1994. Nevertheless, the distinction between these two moments of the vigilance system will be important in understanding the data reviewed here.

This distributed model of visual spatial attention has been attractive to psychopathologists interested in attentional dysfunction, in a wide range of disorders, for several reasons. The interactions among these three attentional networks exemplify the brain's capacity to balance reflexive or data-driven processes and controlled or goal-oriented processes. All three systems are therefore potentially relevant to understanding the core mechanism of dysfunction in children with ADHD. Further, the model features a well-validated conceptual neural system with a portable and well-designed measurement probe that is suitable to behavioral research. This combination of features means that any positive findings may be able to shed light on which neuroanatomical networks are involved in ADHD (or other disorders). It is not surprising then that psychopathologists have been eager to employ this paradigm. We now briefly describe the general measurement approach used across all studies reviewed.

Fig. 1 presents a schematic diagram of the computerized target detection task in which RT is the main outcome measure. Participants first fixate on a cross hair located at the center of a display, flanked on two sides by a box. Then, a cue directs attention to a target (usually an asterisk) that will shortly appear in one of the two boxes. The participant presses a keyboard button on detecting the target.

Valid cues direct attention to the box in which the target will appear and shorten the RT to detect the target (RT “benefit”). Invalid cues direct attention to the box in which the target will not appear and lengthen the RT to correct target detection (RT “cost”). Neutral cues are alerting but do not direct attention to a specific location because both boxes are simultaneously cued (alternatively, “no cue” conditions are sometimes used to isolate alerting itself). In general, the cueing costs (invalid minus neutral RT) are thought to probe the disengage process, the cueing benefits (neutral minus valid RT) involve the move and engage processes, and the validity effect (invalid minus valid RT) provides a measure of the overall strength of cue response. However, the value of cost and benefit analysis has been questioned (Jonides & Mack, 1984).

Moreover, as illustrated in Fig. 1, attention can be directed by two kinds of cues: exogenous or endogenous. Exogenous directional cues appear as a peripheral “brightening” of one box, drawing attention largely automatically (Jonides, 1981) and exemplifying activity of the PAS (Posner & Raichle, 1994). Endogenous directional cues appear as a central arrow pointing to one box, requiring voluntary movement of attention to the target and involving the AAS. Ratio of valid to invalid cues is important. Most (e.g., 80%) endogenous cues are valid, providing participants a strategic incentive to send their attention in the direction of the arrow. Exogenous cues provide the purest measure of the PAS at a 50% validity rate (so there is no strategy benefit). However, these methods of cueing are mixed in various ways as noted later.

Finally, the period between the appearance of the cue and the onset of the target, or stimulus onset asynchrony (SOA), is noteworthy. Short SOAs (e.g., <100 ms for adults) are too rapid for attention to be moved voluntarily (Jonides, 1981) and so are usually used only with exogenous cues. Longer SOAs (e.g., >300 ms) are usually used when participants are endogenously cued. With the theory and task parameters in mind, we now turn to the literature.