Poor Working Memory Starts Early in Processing Pipeline
15 March 2011. Efforts to explain faulty working memory in schizophrenia have focused on prefrontal cortical regions, but a new study fingers abnormal processing in sensory cortex instead. Daniel Javitt of the Nathan Kline Institute for Psychiatric Research in Orangeburg, New York, and colleagues measured event-related potentials in subjects with schizophrenia and in healthy subjects who were performing a task that taxes working memory. The results, published online in the Archives of General Psychiatry on March 7, tie working memory deficits in schizophrenia to reduced activation of the visual cortex, particularly in the low-resolution magnocellular pathway. Most importantly, the activation that best predicted task performance came from sensory rather than cognitive processing regions, adding to evidence that impaired working memory in schizophrenia reflects faulty stimulus encoding rather than just an inability to hold items in memory. These findings challenge the view that prefrontal dysfunction alone underlies working memory impairment in schizophrenia.
A recent meta-analysis found consistent support for large working memory deficits in schizophrenia (see Forbes et al., 2009). To assess these deficits, researchers often use the AX-type continuous performance test (AX-CPT), which tasks subjects with responding to letters that appear one at a time on a computer screen. To respond correctly, subjects must press a button only when the letter X follows the letter A and not after other sequences, collectively called BX or AY. Thus, they must correctly encode the first, or cue, letter, and keep it in memory until they see the second letter, called the probe.
During such tasks, subjects with schizophrenia show altered prefrontal activation compared to healthy subjects, according to functional magnetic resonance imaging studies (see meta-analysis by Minzenberg et al., 2009). In an interview with SRF, Javitt said that researchers have assumed that the working memory deficits in schizophrenia arise in dopamine-rich prefrontal cortex, despite evidence that sensory processing deficits also occur in schizophrenia (see SRF related news story). Furthermore, he noted, administering the N-methyl-D-aspartate receptor blocker ketamine to healthy subjects induces schizophrenia-like performance on the AX-CPT, as might be expected from glutamatergic and other whole-brain theories of schizophrenia (see SRF Current Hypothesis by Javitt).
Javitt said that individual imaging studies have not found decreased activity in visual cortex in schizophrenia, but he noted that they “have never really been optimized to look at the visual areas; they have been designed in such a way that they’re very sensitive to frontal cortex.” Consequently, he and his colleagues, including first author Elisa Dias, also of the Nathan Kline Institute, used event-related potentials, a measure more sensitive to visual processing, to explore whether sensory processing abnormalities can explain visual working memory deficits in schizophrenia. Specifically, they compared AX-CPT performance and corresponding brain activity in 30 patients with schizophrenia, all of whom were receiving treatment with antipsychotic drugs, and 17 healthy subjects.
In order to parse the cognitive processes involved, Dias and colleagues tested each subject with three different versions of the AX-CPT, which differed only in stimulus probabilities. For instance, the AX-70 version presented the AX sequence on 70 percent of trials, whereas in the AY-70, the most probable sequence consisted of an A followed by a letter other than X, and in the BX version, a non-A letter followed by X.
On all versions, patients were more likely than control subjects to fail to press the button when AX appeared. When they mistakenly pressed the button, they were more likely than controls to do so in response to BX than to AY sequences, reflecting a weakened ability to use cues to guide behavior. Yet patients were as capable as control subjects of adjusting their behavior to new stimulus probabilities. Behaviorally, their top-down processing seemed intact, yet their performance remained faulty.
Waves of data
To gauge neural activity in different parts of the processing pathway, Dias and colleagues recorded event-related sensory and cognitive potentials. Specifically, they recorded sensory potentials C1 at 60 to 120 milliseconds after the stimulus, P1 at 75 to 130 milliseconds, and N1 at 100 to 200 milliseconds. At later time points, they recorded cognitive potentials N2 at 220 to 350 milliseconds, and CNV, or contingent negative variation, at 1,200 to 1,250 milliseconds. Peak amplitude served as the outcome measure.
Patients with schizophrenia showed smaller activation of P1 and N1, but not C1, compared to healthy subjects. This shows that they may have deficits in early sensory processing, particularly in the magnocellular visual system, the more dorsal of the two pathways that carry sensory information to prefrontal areas. This system provides low-resolution information to guide action, whereas the parvocellular system supplies finer-honed information for identifying stimuli. The latter system, indexed by the C1, seems relatively intact in schizophrenia. Further tests of subjects’ ability to detect low-contrast stimuli at the spatial frequencies detected by the two systems further pointed to magnocellular processing deficits in schizophrenia.
Not surprisingly, patients with schizophrenia showed dampened activation at higher, cognitive levels of processing. Their N2 and CNV potentials—measures of activity in frontal regions—showed smaller peaks, evidence of impaired encoding of information about the cue.
Bottom-up versus top-down
Regression analyses found that the N1 potentials, which come from visual cortex, best predicted behavior. They correlated significantly with AX-70 performance (r = 0.37, P = .004), overshadowing N2’s marginal ties to performance.
Path analysis hinted that P1 sensory processing might be driving N2 cognitive potentials. This again put magnocellular deficits at the heart of schizophrenia-related impairments. This notion gained further support from significant correlations between task performance and subjects’ ability to detect the kinds of low-contrast stimuli processed by the magnocellular system.
Taken together, Dias and colleagues see their findings as further confirmation of the idea that working memory impairments in schizophrenia reflect failures of encoding rather than of memory retention. They do not dispute that top-down processing breaks down in schizophrenia; rather, they argue that bottom-up processing deficits deserve more attention than they have received. As Javitt said, these findings suggest “that not only are the frontal activation deficits in schizophrenia due to sensory deficits but, in fact, the behavioral deficits are really reflecting the sensory dysfunction more than the higher frontal dysfunction.”
Javitt thinks that such findings point to a need to rethink cognitive remediation in schizophrenia. Instead of directly targeting higher-level processes, such as attention and working memory, focusing on sensory processing might well produce effects that ripple through the brain. “That might not fix everything, but until you fix that, it will be hard for patients to learn and improve other aspects of performance,” he said.—Victoria L. Wilcox.
Dias EC, Butler PD, Hoptman MJ, Javitt DC. Early sensory contributions to contextual encoding deficits in schizophrenia. Arch Gen Psychiatry. 2011 Mar 7. Abstract
Comments on News and Primary Papers
Comment by: Steven Silverstein
Submitted 19 June 2011
Posted 19 June 2011
I recommend the Primary Papers
This new paper by Dias et al. is important in at least two respects. First, it is further evidence of visual processing impairment in schizophrenia. Second, and perhaps more importantly, as noted by the authors, the findings provide more evidence that schizophrenia is not primarily a disorder of cognitive control or prefrontal cortex function. That is, while schizophrenia does involve those impairments, it appears likely that they can be viewed as 1) being affected by inadequate representations emerging from more basic aspects of processing such as perception; and 2) being due to a primary disorder of neural network formation and maintenance—a widespread failure that affects all aspects of cognitive functioning. These points are discussed in detail in a forthcoming theme section of Schizophrenia Bulletin on Vision Science and Schizophrenia Research.
As noted by Javitt in the article summary, the findings by Dias et al. suggest “that not only are the frontal activation deficits in schizophrenia due to sensory deficits, but, in fact, the behavioral deficits are really reflecting the sensory dysfunction more than the higher frontal dysfunction.” While this statement may seem controversial, or even heretical, it is, nevertheless, also supported by data.
For example, the study is consistent with past behavioral, ERP, and fMRI data showing how working memory deficits in schizophrenia are downstream effects of inadequate sensory and perceptual processing (e.g., Haenschel et al., 2007; Silverstein et al., 2005). In fact, only in cases where stimuli were used for which it is known that schizophrenia patients do not have a problem (e.g., basic color perception) has it been found that perceptual factors do not limit working memory function (e.g., Gold et al., 2010).
On the broader point noted above, regarding the benefits of a greater understanding of visual processing impairments in schizophrenia, there are several relevant considerations. One is that many, if not all, of the processes typically studied in schizophrenia can be more easily and clearly studied using paradigms from vision science. For example, context processing (the focus of the study by Dias et al.) occurs in visual cortex as well (at several levels, including gain control, center-surround modulation, motion perception, collinear facilitation, and other effects of spatial context on perception, and effects of task context such as the order in which stimuli or conditions are presented). Moreover, other functions such as re-entry and efference copy/collateral discharge, which are relevant to schizophrenia, can also be studied in very concrete ways in the visual cortex. Studies of all of these phenomena in schizophrenia have found abnormalities suggesting that basic modulatory processes are impaired at the perceptual level, and that this may be the operative factor in "high-level" impairments such as those involving working memory. In support of this is a recent study on cognitive control in schizophrenia that attributed its failure to impaired contextual modulation in the frontal cortex (Barbalat et al., 2009).
Although not noted in the SRF summary of the Dias paper, an important point relevant to their view that sensory and perceptual abnormalities are primary disturbances in schizophrenia is (as noted in the Introductory paper to the forthcoming Schizophrenia Bulletin theme section on Vision Science and Schizophrenia Research): there is also much evidence that 1) many aspects of perception do not require cognitive control; and 2) there are disturbances in perceptual processing in schizophrenia that cannot be explained in terms of higher-level deficits. Regarding the first point, for example, intact contour integration has been observed in the neglected visual field in patients with hemispatial neglect, and in healthy subjects, contour integration happens automatically even when participants do not want it to or are not attending to the contour-containing stimuli. Moreover, visual processes such as spatial frequency processing, perceptual organization, form perception, and motion perception are found throughout the animal kingdom in species with far less frontal cortex development than humans. Regarding the second point, involvement of the occipital lobe in contour integration has been found in anesthetized monkeys, with the same regions showing hypoactivation in studies with schizophrenia patients, suggesting that the basic binding process that is impaired does not normally require cognitive control. Similarly, contour integration deficits have been repeatedly demonstrated in people with amblyopia, a condition involving reduced integration of information in early visual cortex regions, associated with suppressed input from one eye, and these patients are not characterized by impairments in frontal cortex functioning.
Further, as demonstrated by Kéri and colleagues (Kéri et al., 2009), in collinear facilitation tasks, when both schizophrenia patients and controls attend to flankers to the same degree (as verified by a secondary task), only patients show impaired ability at integrating the flanker elements. Patients are also less susceptible to certain illusions, and it is doubtful that deficits in cognitive control help subjects recover correct shape information. In addition, there is abundant evidence for magnocellular pathway deficits in schizophrenia, based on studies of motion, spatial frequency, and contrast sensitivity processing. These deficits have well-known origins in early sensory processing centers. Finally, findings of increased occipital lobe gray matter loss in poor outcome patients (see Mitelman and Buchsbaum, 2007), the subgroup that typically shows the most severe perceptual impairments, suggests a primary role for occipital lobe dysfunction in visual stimulus assembly failures in schizophrenia.
In short, the evidence from vision research highlights that much coordination of perceptual and cognitive activity emerges via self-organization in local populations of neurons, which is a general property of circuitry throughout the brain. This view does not negate the importance of different brain regions for specialized processing, such as that the occipital lobe is a visual information processor, whereas the prefrontal cortex is heavily involved in strategic planning and goal maintenance. Nor does it negate the possibility that there are abnormalities in schizophrenia that are due to faulty interregional interactions. Rather, this view highlights the primary importance of coordinating processes for diverse mental functions, the likelihood that at least some perceptual phenomena and their abnormalities in schizophrenia are best accounted for by failures in this process within visual pathways, and the relative simplicity of viewing and comprehending coordinating processes, and their impairments in schizophrenia, through the lens of vision research.
Finally, a minor correction is necessary to the statement in the SRF summary that studies have not found hypoactivation in visual cortex regions in schizophrenia. In fact, this has been found, in studies of both visual processing (e.g., Silverstein et al., 2009) and working memory (the study by Haenschel et al. noted above). Moreover, in some cases, as with face processing, compensatory activity in higher regions has been observed in the context of reduced activity in striate and/or extrastriate visual cortex regions (e.g., Silverstein et al., 2009; Silverstein et al., 2010).
In short, there is compelling evidence for visual processing dysfunction in schizophrenia (along with impairments in other sensory domains), and these impairments serve as rate limiting factors for later processing. They also demonstrate, in concrete form, much about what is altered throughout the brain in schizophrenia. Greater attention to these issues can have the effect of refining our view of schizophrenia as a whole brain disorder, and moving away from the view that a primary failure in executive control (of presumably normal representations) defines what schizophrenia is.
Haenschel C, Bittner RA, Haertling F, Rotarska-Jagiela A, Maurer K, Singer W, Linden DE. Contribution of impaired early-stage visual processing to working memory dysfunction in adolescents with schizophrenia: a study with event-related potentials and functional magnetic resonance imaging. Arch Gen Psychiatry . 2007 Nov 1 ; 64(11):1229-40. Abstract
Silverstein SM, Bakshi S, Nuernberger S, Carpinello K, Wilkniss S. Effects of stimulus structure and target-distracter similarity on the development of visual memory representations in schizophrenia. Cogn Neuropsychiatry . 2005 Jun 1 ; 10(3):215-29. Abstract
Gold JM, Hahn B, Zhang WW, Robinson BM, Kappenman ES, Beck VM, Luck SJ. Reduced capacity but spared precision and maintenance of working memory representations in schizophrenia. Arch Gen Psychiatry . 2010 Jun 1 ; 67(6):570-7. Abstract
Barbalat G, Chambon V, Franck N, Koechlin E, Farrer C. Organization of cognitive control within the lateral prefrontal cortex in schizophrenia. Arch Gen Psychiatry . 2009 Apr 1 ; 66(4):377-86. Abstract
Kéri S, Kelemen O, Benedek G. Attentional modulation of perceptual organisation in schizophrenia. Cogn Neuropsychiatry . 2009 Mar 1 ; 14(2):77-86. Abstract
Mitelman SA, Buchsbaum MS. Very poor outcome schizophrenia: clinical and neuroimaging aspects. Int Rev Psychiatry . 2007 Aug 1 ; 19(4):345-57. Abstract
Silverstein SM, Berten S, Essex B, Kovács I, Susmaras T, Little DM. An fMRI examination of visual integration in schizophrenia. J Integr Neurosci . 2009 Jun 1 ; 8(2):175-202. Abstract
Silverstein SM, All SD, Kasi R, Berten S, Essex B, Lathrop KL, Little DM. Increased fusiform area activation in schizophrenia during processing of spatial frequency-degraded faces, as revealed by fMRI. Psychol Med . 2010 Jul 1 ; 40(7):1159-69. Abstract
View all comments by Steven SilversteinComment by: Michael F. Green
Submitted 30 June 2011
Posted 30 June 2011
The paper by Dias et al. makes an important contribution to our understanding of perceptual processes in schizophrenia. Steve Silverstein has already provided a thoughtful and detailed comment on it. I have just three points to add:
First, the authors provide an excellent demonstration of the explanatory value of bottom-up perceptual processing for vigilance and working memory tasks, such as the AX version of the continuous performance task. However, the implications extend beyond such tasks. There is increasing support for the importance of early perceptual processes (both auditory and visual) for social cognition, including prosody detection and social perception (Leitman et al., 2005; Wynn et al., 2010). In addition, early visual processing has been part of outcome models that reach to community functioning (Rassovsky et al., 2011; Butler et al., 2005). Therefore, the value of perceptual models is that they influence how we interpret cognitive tasks that have typically been viewed in terms of top-down control processes (as in Dias et al.). They also provide a way to map the steps from brain processes to daily functioning.
Second, when it comes to mapping the stages of processing, EEG is still king. Although functional magnetic resonance imaging (fMRI) is increasingly used to examine briefly presented stimuli, there is still no better way than EEG to separate the cortical response to one stimulus versus another, or to separate stages of processing. This is particularly true when one is trying to distinguish among stages of perceptual processing, or between sensory and cognitive event-related potentials.
Third, it should be noted that a bottom-up focus (as described in Dias et al.) does not necessarily implicate the earliest stages of perceptual processes. There are several examples in which patients and controls differ at early, but not the earliest, processing stages. Nor do the earliest stages always account for variance in later stages. For example, we have consistently found visual backward masking deficits in schizophrenia after matching subjects on their accuracy for detecting an unmasked target. In other words, the ability to detect a briefly presented target alone does not explain patients’ problems in processing visual stimuli presented in rapid succession (Green et al., 2003). Also, we have found that fMRI activation in primary visual areas (i.e., retinotopic areas) is intact in schizophrenia, but that differences emerge at slightly later visual stages, such as object perception, which is conducted by the lateral occipital complex (Green et al., 2009).
Lastly, we have found that ERP (event-related potential) components of emotional responses to evocative pictures are intact for the first few hundred milliseconds following stimulus onset (Horan et al., 2010).
The implication is that bottom-up models have terrific explanatory value for higher-level cognitive processes, social cognition, and even community functioning. At the same time, they have their limits and present their own mysteries, such as why bottom-up does not always seem to start at the bottom.
Leitman DI, Foxe JJ, Butler PD, Saperstein A, Revheim N, Javitt DC. Sensory contributions to impaired prosodic processing in schizophrenia. Biol Psychiatry. 2005;58(1):56-61. Abstract
Wynn JK, Sugar C, Horan WP, Kern R, Green MF. Mismatch negativity, social cognition, and functioning in schizophrenia patients. Biol Psychiatry. 2010;67(10):940-7. Abstract
Rassovsky Y, Horan WP, Lee J, Sergi MJ, Green MF. Pathways between early visual processing and functional outcome in schizophrenia. Psychol Med. 2011;41:487-97. Abstract
Butler PD, Zemon V, Schechter I, Saperstein AM, Hoptman MJ, Lim KO, Revheim N, Silipo G, Javitt DC. Early-stage visual processing and cortical amplification deficits in schizophrenia. Arch Gen Psychiatry. 2005;62(5):495-504. Abstract
Green MF, Nuechterlein KH, Breitmeyer B, Tsuang J, Mintz J. Forward and backward visual masking in schizophrenia: influence of age. Psychol Med. 2003;33:887-95. Abstract
Green MF, Lee J, Cohen MS, Engel SA, Korb AS, Nuechterlein KH, Wynn JK, Glahn DC. Functional neuroanatomy of visual masking deficits in schizophrenia. Arch Gen Psychiatry. 2009;66(12):1295-1303. Abstract
Horan WP, Wynn JK, Kring AM, Simons RF, Green MF. Electrophysiological correlates of emotional responding in schizophrenia. J Abnorm Psychol. 2010;119(1):18-30. Abstract
View all comments by Michael F. Green