ARCHIVAL REPORT

Working Memory in Schizophrenia: Behavioral and Neural Evidence for Reduced Susceptibility to Item-Specific Proactive Interference Christoph P. Kaller, Sandra V. Loosli, Benjamin Rahm, Astrid Gössel, Stephan Schieting, Tobias Hornig, Jürgen Hennig, Ludger Tebartz van Elst, Cornelius Weiller, and Michael Katzev Background: Susceptibility to item-specific proactive interference (PI) contributes to interindividual differences in working memory (WM) capacity and complex cognition relying on WM. Although WM deficits are a well-recognized impairment in schizophrenia, the underlying pathophysiological effects on specific WM control functions, such as the ability to resist item-specific PI, remain unknown. Moreover, opposing hypotheses on increased versus reduced PI susceptibility in schizophrenia are both justifiable by the extant literature. Methods: To provide first insights into the behavioral and neural correlates of PI-related WM control in schizophrenia, a functional magnetic resonance imaging experiment was conducted in a sample of 20 patients and 20 well-matched control subjects. Demands on item-specific PI were experimentally manipulated in a recent-probes task (three runs, 64 trials each) requiring subjects to encode and maintain a set of four target items per trial. Results: Compared with healthy control subjects, schizophrenia patients showed a significantly reduced PI susceptibility in both accuracy and latency measures. Notably, reduced PI susceptibility in schizophrenia was not associated with overall WM impairments and thus constituted an independent phenomenon. In addition, PI-related activations in inferior frontal gyrus and anterior insula, typically assumed to support PI resistance, were reduced in schizophrenia, thus ruling out increased neural efforts as a potential cause of the patients’ reduced PI susceptibility. Conclusions: The present study provides first evidence for a diminished vulnerability of schizophrenia patients to item-specific PI, which is presumably a consequence of the patients’ more efficient clearing of previously relevant WM traces and the accordingly reduced likelihood for item-specific PI to occur.

Key Words: Cognitive control, functional neuroimaging, prefrontal cortex, proactive interference, schizophrenia, working memory

W

orking memory (WM) impairments constitute a core deficit of cognition in schizophrenia (1,2), but remarkably little is known about the pathological effects on specific WM control processes. Current theories of WM stress mechanisms of interference control as an important source of interindividual variations in WM performance (3,4), which are also predictive for a broad range of other cognitive abilities (5–9). Item-specific proactive interference (PI) in WM occurs if processing of task-relevant information is influenced by lingering representations of previously activated but currently irrelevant From the Department of Neurology (CPK, SVL, AG, TH, CW, MK), University Medical Center Freiburg, Freiburg; Freiburg Brain Imaging Center (CPK, SVL, AG, JH, LTvE, CW, MK), and Biological and Personality Psychology (SVL), Department of Psychology, University of Freiburg, Freiburg; Medical Psychology and Medical Sociology (BR), University Medical Center Mainz, Mainz; Center for Psychiatry (SS), Emmendingen; Department of Psychiatry (TH, LTvE), and Medical Physics (JH), Department of Radiology, University Medical Center Freiburg, Freiburg; and BrainLinks-BrainTools Cluster of Excellence (CPK, JH, CW), University of Freiburg, Freiburg, Germany. Address correspondence to Christoph Kaller, Ph.D., University Medical Center Freiburg, Department of Neurology, Breisacher Str. 64, Freiburg 79106, Germany; E-mail: [email protected]. Received Aug 15, 2013; revised Mar 7, 2014; accepted Mar 11, 2014.

0006-3223/$36.00 http://dx.doi.org/10.1016/j.biopsych.2014.03.012

WM contents (10). Efficient control mechanisms for protecting WM from item-specific PI may therefore be regarded as an important prerequisite for sustaining coherent lines of thought and action. The Cognitive Neuroscience Treatment Research to Improve Cognition in Schizophrenia consortium (11) suggested the recentprobes task (12) (Figure 1A) as a promising paradigm to investigate PI susceptibility in patients with schizophrenia (PSZ) (13,14). In healthy control subjects (HCS), reduced WM performance is associated with increased PI susceptibility (3,15). Given their well-documented WM deficits (2), PSZ might thus be expected to exhibit an increased vulnerability to PI effects. A similar expectation might be derived from studies suggesting a specific deficit of PSZ in the controlled selection from WM (16). However, item-specific PI can only arise from sufficiently stable representations of information in WM. If PSZ showed substantial deficits in WM performance, item-specific PI might therefore be less likely to arise from (unreliably maintained) previous WM contents. That is, assuming an inverted U-shaped relationship between WM performance and PI susceptibility (Figure 1B), the opposite prediction of a reduced PI susceptibility in schizophrenia would also be conceivable. Besides the dependency of PI susceptibility on WM performance, demands on interference control might also be otherwise affected in schizophrenia. Theoretical frameworks on schizophrenia postulating a reduced influence of past experience on current information processing (17,18) would predict a reduced PI susceptibility in PSZ compared with HCS. However, as other aspects of inhibition-related control, such as restraining prepotent response tendencies, are known to be impaired in schizophrenia (19), one might also derive the opposite expectation that PSZ show an increased susceptibility BIOL PSYCHIATRY 2014;]:]]]–]]] & 2014 Society of Biological Psychiatry

C.P. Kaller et al.

2 BIOL PSYCHIATRY 2014;]:]]]–]]] Trial n-1

Association between PI Susceptibility and WM

Proactive Interference

b‘

b c a b“

Trial n Working Memory (WM) Accuracy

Recent-probes variant of the Sternberg Item Recognition Task

...

Figure 1. (A) Illustration of the recent-probes variant (12) of the Sternberg Item Recognition task (32). The Sternberg Item Recognition Task constitutes a delayed match-to-sample working memory (WM) paradigm. Target sets of items (here, four pictures of to-be-encoded animals) are presented for a short period of time after which they disappear. Following a delay, a probe item is presented requiring the subject to decide whether it was part of the target set or not. Commonly, one half of trials consist of positive probes matching one of the targets, whereas the other half of trials comprise negative nonmatching probes (factor probe type). In the recent-probes variant, matching of the probe is manipulated not only with respect to the current trial (n) but also with respect to the preceding trial (n-1) (10,12). In recent trials, the probe matches one of the targets in the preceding trial, whereas it does not match one of the targets in nonrecent trials (factor recency). The two factors probe type and recency are independently varied in a 2
2 factorial design. Item-specific proactive interference (PI) is specifically elicited in recent negative trials in which the probe was present as a target in the preceding but not the current trial, leading to false-positive decisions and increased response latencies (10). As illustrated, target and probe items in the present study were drawn from a pool of 12 achromatic line drawings of animals from the Snodgrass and Vanderwart (63) picture set (reprinted with permission, American Psychological Association). All animal pictures could be named with monosyllable German words. (B) Schematic illustration of the inverted U-shaped relationship between WM performance and PI susceptibility. In general, the occurrence of item-specific PI is highly dependent on the reliability of information representation in WM. Optimal WM representations, which are highly stably maintained representations of the present trial’s target items, are unlikely to be confused with those from the previous trial when matched against the probe item. Likewise, if WM representation is at chance level, representations of the target items are too unstable and erratic and thus cannot interfere with the probe decision of the subsequent trial. In between these two extremes, item-specific PI inevitably arises as a function of WM performance, given that more or less reliably maintained representations of target items from the present trial compete with more or less reliably maintained representations of target items from the previous trial. That is, for moving from either extreme toward the other (WM representation at optimum vs. at chance), PI susceptibility consistently increases and eventually reaches its peak at a point where the stabilities of the competing representations of target sets trade off. Thus, the negative association with PI susceptibility reported for samples with normal to higher levels of WM performance (3,15) only partially reflects the actual bilinear/quadratic relationship, as a positive association with PI susceptibility necessarily results at lower levels of WM performance. As exemplified with the representative black dots on the solid line, schizophrenic patients compared with matched control subjects can thus be expected to have either an increased (e.g., dots b vs. a) or a decreased (c vs. b) or even a comparable susceptibility to item-specific PI (c vs. a), depending on the samples’ respective positions on the inverted U-shape function. In addition, predictions become yet more complicated if one takes into account that the inverted U-shape function does not necessarily have an identical slope in patients and control subjects. For instance, the (passive) decay and/or (active) clearance of WM representations are other highly relevant factors influencing PI susceptibility across trials. For instance, slower versus faster decay should result in flatter versus steeper curvilinear relations (e.g., gray dots b’ and b” on dashed lines).

to PI by responding impulsively to familiar stimuli. This assumption might be further supported by findings in healthy subjects that a higher susceptibility to PI is related to a higher incidence of unwanted intrusive thoughts (20). Taken together, competing predictions of an increased versus a decreased PI susceptibility can be likewise postulated for WM www.sobp.org/journal

performance of PSZ in the recent-probes task. Thus far, no empirical data are available to decide between these alternative hypotheses. The present study was set out to provide first insights into PI-related WM control functions in schizophrenia and to elucidate their neural underpinnings. Note that, unlike the processes of item-specific PI addressed here, effects of item-

C.P. Kaller et al.

BIOL PSYCHIATRY 2014;]:]]]–]]] 3

Table 1. General Demographic Information and Clinical Characteristics Normal Control Subjects (n ¼ 20) Mean Age (Years) Sex Crystallized Intelligence (MWT-B) (31) Duration of Illness (Years) SANS (25) SAPS (24) TLC (26) HAM-D (27) YMRS (28)

30.7 28.7

SD

Min

10.0 18.5 11 male/9 female subjects 4.0 19

Schizophrenic Patients (n ¼ 20) Max

Mean

47.4

30.8

35

27.5 6 28.60 28.55 10.10 7.84 3.58

SD

Min

10.7 18.3 11 male/9 female patients 5.1 18 6.3 1 16.14 1 20.83 1 10.67 0 5.96 0 3.15 0

Max 48.8 36 21 53 67 40 19 11

HAM-D, Hamilton Rating Scale for Depression; Max, maximum; Min, minimum; MWT-B, Mehrfachwahl-Wortschatztest; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms; TLC, Scale for the Assessment of Thought, Language, and Communication; YMRS, Young Mania Rating Scale.

unspecific PI have been previously investigated in schizophrenia, partly demonstrating impaired release from item-unspecific PI in PSZ [e.g., (21,22)]. But despite the terminological overlap, itemspecific and item-unspecific PI relate to different cognitive phenomena.1

Methods and Materials Subjects In total, 23 PSZ (mean age ⫾ SD, 30.6 ⫾ 9.9 years; 14 male subjects; all right-handed) participated in the present experiment. Patients were recruited at inpatient and outpatient facilities of the Department of Psychiatry of the University Medical Center Freiburg and the Center for Psychiatry Emmendingen. All PSZ were diagnosed with schizophrenia or schizoaffective disorder according to the DSM-IV criteria (23). All but one of the PSZ were administered either monotherapy or multiple-drug therapy with stable doses of atypical antipsychotics (Supplement 1), which was combined with chlorprothixene in one patient. One patient received monotherapy with flupenthixol. Diagnoses were complemented by a neurological examination, a battery of neuropsychological tests, a clinical electroencephalogram, and a screening examination using anatomical magnetic resonance imaging. Exclusion criteria were inability to give informed consent, a history of substance abuse within the past 6 months, or a history of significant neurological disorders. Psychopathological symptoms were assessed in a clinical interview by a specialist in psychiatry using the Scale for the Assessment of Positive Symptoms (24), the Scale for the Assessment of Negative Symptoms (25), the Scale for the Assessment of Thought, Language, and Communication (26), the Hamilton Rating Scale for Depression (27), and the Young Mania Rating Scale (28). Data of control subjects were drawn from a pool of 47 datasets of right-handed HCS that all had accomplished the same experimental task under the same conditions in the same magnetic resonance scanner as the PSZ. Mental health of HCS 1

Item-unspecific PI refers to interference from accumulating (semantically related) information that has been remembered in the course of the task, so that recall typically declines across succeeding trials (and is released for a semantically unrelated category). In contrast, itemspecific PI refers to interference specifically induced by the familiarity of the recent negative probe item of the current trial, given that it has been a target item in the immediately preceding trial but is not in the present trial.

was screened using the Mini International Neuropsychiatric Interview (German Version 5.0.0) (29) and the Structural Clinical Interview for DSM-IV (German Version) (30) for Axis I and Axis II disorders, respectively. Inspection of the behavioral data indicated that three PSZ did not conform to the task by showing a strong response bias (Supplement 1). These PSZ were excluded from all subsequent analyses. For pair-wise matching, a multidimensional matching approach was applied (Supplement 1) to account for betweengroup differences in age, sex, and premorbid verbal crystallized intelligence. The latter was assessed with a German vocabulary test (Mehrfachwahl-Wortschatztest) (31). In consequence, present analyses comprised 20 well-matched pairs of PSZ and HCS. An overview of the final sample’s demographic and clinical characteristics is provided in Table 1. The study was approved by the Ethics Committee of the University of Freiburg (votum #197/10) and conducted in accordance with the guidelines of the World Medical Association Declaration of Helsinki (http://www.wma.net). Informed consent was obtained before participation. All participants received compensation of €50 for participation. Experimental Task and Design Item-specific PI was experimentally manipulated in a recentprobes task (12), which is a variant of the Sternberg Item Recognition task (32). In each single trial, first a set of four target items was presented for encoding (2000 msec) followed by a delay interval (2400 msec). Thereafter, a probe item was presented (1800 msec) that prompted the participants to decide whether it was part of the target set (Figure 1). Positive or negative answers were given by pressing a corresponding button on a magnetic resonance-compatible computer mouse. The task comprised a 2
2 factorial design: the factor probe type determined whether the probe item was part of the current target set n (positive vs. negative), whereas the factor recency manipulated whether it was part of the target set of the immediately preceding trial n-1 (recent vs. nonrecent) (Figure 1). In the recent-probes task, PI is elicited in trials with a recent negative probe, i.e., in which the probe was part of the target set in the preceding but not in the current trial. In these trials, still active but now irrelevant memory traces from the previous target set interfere with the currently relevant information in WM, thus leading to decreased response accuracies in terms of false-positive decisions and increased response latencies (10). www.sobp.org/journal

C.P. Kaller et al.

4 BIOL PSYCHIATRY 2014;]:]]]–]]] The experiment consisted of three runs with 64 trials each. The four conditions of the 2
2 design were equally distributed within runs (16 trials per run). To prevent a confound between item-specific PI and item familiarity, pseudorandom selection of target items across trials ensured that in every trial, half of the target items were already presented as targets (but not as probe) in the immediately preceding trial, whereas the other half of the target items had not occurred as targets in the preceding three trials (33). Thus, because all target sets comprised an equal number of directly repeated and nonrepeated items, any potential cognitive and neural effects of probe type and recency were exclusively restricted to the presentation of the probe item (33). Imaging Procedure and Analysis For detailed information on the imaging procedure and parameters of the T2*-weighted echo-planar imaging sequence, data preprocessing, and first-level analyses, please refer to Supplement 1. First-level estimates of hemodynamic activation changes were computed based on the general linear model implemented in SPM8 (Wellcome Trust Centre for Neuroimaging, University College London, United Kingdom; http://www.fil.ion.ucl.ac.uk/ spm/software/spm8). Encoding and delay were modeled by a single regressor independent of conditions, whereas responses to the probe were modeled by individual regressors for each of the four conditions (33). As the focus of present analyses was on itemspecific PI, normalized beta images for the recent and nonrecent negative conditions were entered into second-level analyses of variance using GLM FLEX (Aaron P. Schultz, Martinos Center for Biomedical Imaging at the Massachusetts General Hospital, Charlestown, Massachusetts; http://nmr.mgh.harvard.edu/har vardagingbrain/People/AaronSchultz/GLM_Flex.html; release 2/ 28/12). Second-level analyses comprised a flexible factorial model accounting for effects of subjects, as well as of the two main effects of interest (group, recency) and their interaction. Analyses followed a two-step procedure. First, PI-related neural activation was identified on the whole-brain level and across groups (contrasting recent vs. nonrecent trials). To this end, significance of activations was corrected for multiple comparisons on the voxel level by using a false-discovery rate (FDR) of p ⬍ .05 and assessed at a cluster extent of k ⬎ 50 voxels (.169 cm3). In a second step, group and interaction effects were assessed within the PI-related network using volume-of-interest (VOI) analyses with a spherical VOI centered on the major peaks for the groupindependent main effects of PI. Volume-of-interest data were extracted with a radius of 6 mm and aggregated based on the first eigenvariate. By using independent contrasts for identifying PI-relevant brain structures and for assessing differences in PI-related neural activation between groups in terms of the interaction effect, inflated statistics due to double-dipping were effectively avoided as differences were not exaggerated (34).

Results Behavioral Markers of Item-Specific Proactive Interference Behavioral data in the recent-probes task were analyzed in two steps. First, global differences in WM performance were assessed in nonrecent positive and negative trials, that is, irrespective of potentially different PI susceptibilities. PI was analyzed in a second step comprising recent negative versus nonrecent negative trials, as PI is only elicited by recent negative probes. www.sobp.org/journal

Working Memory (Non-Recent Trials)

Proactive Interference (Negative Trials)

Figure 2. Illustration of (A) the global impairments of schizophrenic patients in working memory (main effect of group) and (B) their reduced susceptibility to item-specific proactive interference (interaction of group and recency). Effects were evident in both accuracy and latency measures (left and right panels, respectively). Bars denote the standard error of mean. *p ⬍ .05; **p ⬍ .01. Neg, negative; Pos, positive; Rec, recent; ~Rec, nonrecent.

Assessing global WM performance in nonrecent trials, a repeated-measures analysis of variance (RM-ANOVA) on response accuracy revealed significant main effects of group (PSZ vs. HCS, F1,38 ¼ 9.170, p ¼ .004, ηp 2 ¼ .194) and probe type (positive vs. negative, F1,38 ¼ 46.133, p ⬍ .001, ηp 2 ¼ .548) but no interaction (F1,38 ¼ .793, p ¼ .379, ηp 2 ¼ .020). PSZ showed a generally lower WM performance compared with HCS, and negative trials were easier than positive trials (Figure 2A). An RM-ANOVA on response latency also yielded significant main effects of group (F1,38 ¼ 4.628, p ¼ .038, ηp 2 ¼ .109), probe type (F1,38 ¼ 4.386, p ¼ .043, ηp 2 ¼ .103), and no interaction (F1,38 ¼ .500, p ¼ .484, ηp 2 ¼ .013). PSZ responded slower than HCS and response latencies in negative trials were longer than in positive trials (Figure 2A). Specifically addressing PI in negative trials, an RM-ANOVA on response accuracy revealed significant main effects of group (F1,38 ¼ 4.373, p ¼ .043, ηp 2 ¼ .103) and recency (F1,38 ¼ 48.431, p ⬍ .001, ηp 2 ¼ .560), as well as a significant interaction (F1,38 ¼ 4.585, p ¼ .039, ηp 2 ¼ .108). An RM-ANOVA on response latency yielded a significant main effect of recency (F1,38 ¼ 58.004, p ⬍ .001, ηp 2 ¼ .604) and a significant interaction (F1,38 ¼ 11.007, p ¼ .002, ηp 2 ¼ .225) but no main effect of group (F1,38 ¼ 2.409, p ¼ .129, ηp 2 ¼ .060). PSZ showed significantly reduced recency effects in both accuracy and latency measures (Figure 2B). Taken together, besides general deteriorations in WM, PSZ were less susceptible to PI than HCS. Neural Responses to Item-Specific Proactive Interference Activation differences elicited by item-specific PI were assessed across groups by contrasting neural responses to recent versus nonrecent negative probes. First, in a whole-brain analysis, PI-

C.P. Kaller et al.

BIOL PSYCHIATRY 2014;]:]]]–]]] 5

+30

+27

+24

+21

-57

-60

A

B

t value

+18

+15

+12

+9

-63

0

1

2

3

4

5

6

Figure 3. Neural correlates for resolving item-specific proactive interference in frontal (A) and parietal (B) cortices. Coronal slices illustrate the spatial distribution of significantly stronger activation following recent versus nonrecent negative probes. Reported activation passed a voxel-wise threshold of p ⬍ .05 (false-discovery rate corrected for multiple comparisons, t ⬎ 2.91). Activations are displayed on the average of the sample’s anatomical scans normalized into the Montreal Neurological Institute stereotactic standard space. Numbers denote slice coordinates in anterior-posterior direction (y).

associated activations were identified irrespective of groups. Besides left inferior parietal sulcus, results revealed a mainly bilateral frontal PI network comprising the inferior frontal gyrus, anterior insula, and dorsomedial prefrontal cortex including the anterior cingulate cortex and supplementary motor area (Figure 3; Table 2). The neural correlates underlying the differential susceptibility to PI in the two groups were assessed in subsequent VOI analyses with spherical seeds (radius 6 mm) that focused on the PI network as defined by the main peaks of the six activation clusters reported in Table 2. Given the behavioral finding that PSZ were less susceptible to item-specific PI, analyzing the underlying neural activation patterns may thus provide valuable insights on whether this was due to the patients’ increased recruitment of neural areas known to resolve PI or, alternatively, whether PI was elicited in PSZ to a lesser extent, accordingly entailing a reduced recruitment of the PI network. Repeated-measures analyses of variance with the factors group (PSZ vs. HCS) and recency (recent vs. nonrecent) were FDR-corrected for multiple comparisons and yielded significant interaction effects in right inferior frontal gyrus (F1,38 ¼ 6.582, pFDR ¼ .042) and left (F1,38 ¼ 10.459, pFDR ¼ .018)

and right anterior insula (F1,38 ¼ 5.643, pFDR ¼ .046) but not in left inferior frontal gyrus or in any of the remaining VOIs (highest F ¼ .547, lowest pFDR ¼ .696). The main effect of group did not reach significance in any of the VOIs (highest F ¼ 3.217, lowest pFDR ¼ .486). As illustrated in Figure 4, PI-related activation was considerably diminished in PSZ compared with HCS. To rule out the possibility that the present VOI analyses may have missed a secondary recruitment of neural resources in the PSZ different from the neural network for resolving PI, we further explored the overall activation patterns induced by recent and nonrecent negative probes as well as their differences in additional wholebrain analyses (Supplement 1). These analyses confirmed that the behavioral finding of PSZ being better (or less affected by PI) than healthy subjects cannot be explained by increased neural efforts, thus indicating that PI is less strongly elicited in the PSZ. Linking Working Memory and Proactive Interference In contrast to previous findings in healthy subjects suggesting an increase of PI effects with poorer WM performance (3,15), PSZ showed a reduced PI susceptibility together with impaired WM

Table 2. Neural Correlates for Resolving Item-Specific Proactive Interference Peak Coordinates (MNI) Activation Cluster

k

Left IFG

1570

Right IFG

1606

Left Anterior Insula Right Anterior Insula ACC, Medial PFC

Left IPS

303 440 1960

219

x 42.0 36.0 49.5 45.0 42.0 30.0 33.0 31.5 10.5 12.0 1.5 9.0 7.5 19.5 30.0 31.5

y 21.0 7.5 16.5 30.5 18.0 19.5 22.5 16.5 10.5 21.0 24.0 15.0 30.0 9.0 6.0 58.5

Statistics z

t

punc

22.5 27.0 24.0 22.5 15.0 4.5 4.5 9.0 46.5 33.0 48.0 51.0 43.5 45.0 42.0 40.5

5.57 5.21 4.85 4.80 4.14 5.44 5.05 3.70 4.94 4.90 4.73 4.60 4.08 3.89 3.59 4.42

⬍105 ⬍105 ⬍104 ⬍104 ⬍104 ⬍105 ⬍105 ⬍103 ⬍105 ⬍105 ⬍104 ⬍104 ⬍103 ⬍103 ⬍103 ⬍104

Reported activation passed a voxel-wise threshold of p ⬍ .05 (false-discovery rate corrected for multiple comparisons, t ⬎ 2.91) and a cluster extent of k ⬎ 50 voxel ( .17 cm3). ACC, anterior cingulate cortex; IFG, inferior frontal gyrus; IPS, inferior parietal sulcus; k, cluster size/number of voxels; MNI, Montreal Neurologic Institute stereotactic standard space; PFC, prefrontal cortex; punc, uncorrected p value.

www.sobp.org/journal

C.P. Kaller et al.

6 BIOL PSYCHIATRY 2014;]:]]]–]]]

t value 0

1

2

3

4

5

6

Figure 4. Volume-of-interest analyses within the neural network for resolving item-specific proactive interference. Axial slices illustrate the main effect of recency, in the upper panel for the triangular parts of left and right inferior frontal gyrus (IFG) and in the lower panel for the left and right anterior insula (see also Figure 3). Bar plots with significant interaction effects indicate that for the respective volumes of interest, patients showed a reduced proactive interference effect compared with normal control subjects. Bars denote the standard error of mean. *p ⬍ .05; **p ⬍ .01. Rec, recent; ~Rec, nonrecent.

performance. Subsequent exploratory analyses therefore focused on the relation between WM and PI in the two groups. Samples of PSZ and HCS were each divided into two subgroups with high and low levels of WM by median split of achieved accuracies in nonrecent trials (Figure 2A). In the HCS, an RM-ANOVA on response accuracy in negative trials with factors WM performance (high vs. low WM) and recency (recent vs. nonrecent) revealed a main effect of recency (F1,38 ¼ 59.931, p ⬍ .001, ηp 2 ¼ .769), a marginally significant main effect of WM performance (F1,38 ¼ 3.289, p ¼ .086, ηp 2 ¼ .154), and most importantly, a significant interaction (F1,38 ¼ 5.184, p ¼ .035, ηp 2 ¼ .224). As evident from Figure 5A, HCS with lower WM performance exhibited stronger susceptibility to PI. Conducting the same analysis in the PSZ yielded significant main effects of WM performance (F1,38 ¼ 28.713, p ⬍ .001, ηp 2 ¼ .615) and recency (F1,38 ¼ 9.507, p ¼ .006, ηp 2 ¼ .346) but no interaction (F1,38 ¼ .005, p ¼ .946, ηp 2 ⬍ .001). Thus, in PSZ, variations in WM performance were not associated with different susceptibility to PI (Figure 5B). However, PSZ and HCS with high WM levels showed a comparable overall performance (t18 ¼ .736, p ¼ .471), whereas PSZ with low WM levels scored substantially worse than HCS with low WM levels (t18 ¼ 5.837, p ⬍ .001; Figure 5). Further, given the presumed inverted U-shaped relationship between WM performance and PI susceptibility (Figure 1B), the missing relation between WM and PI in PSZ might have been artificially introduced by median splitting the patients’ considerably broader distribution of WM performance, potentially collapsing the ascending and descending flanks of the inverted U. To rule out this possibility, samples were subdivided into quartiles. As illustrated in Figure 5C, susceptibility to PI was remarkably reduced in PSZ and, in contrast to HCS, indeed much less related to WM performance. Supplementary analyses in a subsample of PSZ and HCS matched for global WM performance in nonrecent trials (as well as age, sex, and premorbid intelligence) further confirmed that the significantly reduced PI susceptibility in PSZ www.sobp.org/journal

was independent of between-group differences in WM performance (Supplement 1). Correlations of Symptom Complexes with Proactive Interference Further exploratory analyses concerned the question whether reduced susceptibility to item-specific PI in PSZ was associated with specific symptom complexes as assessed with the Scale for the Assessment of Negative Symptoms and Scale for the Assessment of Positive Symptoms questionnaires (24,25). However, none of the applied classification schemes revealed any significant associations with the patients’ reduced susceptibility to PI. Detailed information on symptom distributions, classifications, and subsequent analyses are provided in Supplement 1.

Discussion Present analyses revealed a reduced susceptibility to itemspecific PI in schizophrenia, evident both at the behavioral level (Figure 2B) and the neural level (Figure 4). Most remarkably, this result was not mediated by the patients’ generally lower performance in global WM (Figure 5), as confirmed by additional analyses in a subsample of PSZ and HCS matched for global WM performance (Supplement 1). Thus, the inverted U-shaped relationship between WM performance and PI (Figure 1B) cannot account for the rather paradoxical finding of PSZ being better (or less affected by PI) than healthy subjects. In consequence, patients’ reduced PI susceptibility does not simply reflect those of HCS with low WM performance but indicates qualitative differences in the underlying cognitive and neural processes (35). The ability to resist item-specific PI is reckoned to rely on active control processes that resolve the conflict emerging in recent negative trials from 1) the probe’s high familiarity due to its recent appearance as a target in the previous trial and 2) the

C.P. Kaller et al.

Figure 5. Relation between working memory (WM) and item-specific proactive interference (PI) in (A) control subjects and (B) patients. Groups were each subdivided into two subgroups with high and low WM performance based on a median split of WM accuracy in nonrecent trials (cf. Figure 2A). Thus, WM performance was independently assessed from item-specific PI. Bars denote the standard error of mean. In healthy control subjects, PI susceptibility depended on WM performance with stronger PI effects in subjects with lower WM levels, whereas no such interaction was evident in schizophrenic patients. (C) Assessing the putative nonlinearity in the relationship between WM accuracy and susceptibility to itemspecific PI. For illustrative purposes, groups of patients and control subjects were divided into subgroups based on the quartiles (each comprising n ¼ 5 subjects) of their WM accuracy (see above). Black dots denote quartiles of normal control subjects; white dots indicate patient quartiles. Relations between WM and PI were fitted using a bilinear/ quadratic function assuming that item-specific PI was nonexistent for WM performance at perfect (100%) and chance (50%) levels. *p ⬍ .05. Rec, recent; ~Rec, nonrecent.

absence of any contextual tags that identify the probe as member of the current target set (10). As corroborated here, the resulting competition in classifying the probe as matching or nonmatching is actively resolved by the interplay of several heteromodal brain regions [Figure 3; for review, see (10)], including the inferior frontal gyrus (33,36–39), anterior insula (40), anterior cingulate cortex (40,41), and inferior parietal sulcus (15,39,40). However, the biased competition between recency-based familiarity and lacking of current contextual cues and the resultant response conflict do not seem to arise in PSZ (or at least to a lesser extent). That is, the patients’ reduced susceptibility to item-specific PI in terms of response accuracy entails neither an increase in response latency (Figure 2B) nor an enhanced activity in the neural network underlying PI resolution (Figure 4), both of which would be indicative of a recruitment of additional cognitive and neural resources for (superior) interference resolution. Conversely, PI-related effects in response latency and neural activity were even significantly diminished in PSZ.

BIOL PSYCHIATRY 2014;]:]]]–]]] 7 Thus, their outperformance of HCS in resisting item-specific PI does not seem to be due to enhanced mechanisms of cognitive control but most likely the result of a reduced necessity to apply these mechanisms to protect WM operations from PI. Less demands to protect WM from item-specific PI in trials with recent-negative probes may originate from two different sources of bias toward a correct rejection: 1) a stronger certainty in the probe’s absence of contextual tags from the current target set; and/or 2) a weaker familiarity from the probe’s presence in the recent target set (42,43). Given the consistently reported detriments in WM performance in schizophrenia (2) (Figure 2A), the possibility of a more reliable encoding of the current target set and related contextual information (spatial positions, associations between items, temporal codes, etc.) (10) seems unlikely, thus leaving a diminished influence of recency-based familiarity as the potential cause for the patients’ reduced susceptibility to PI. Conceiving WM as executive attention in terms of traces of long-term memory representations being activated above threshold (44,45), the familiarity of previously relevant information (inducing PI) may therefore be attenuated in schizophrenia due to an accelerated decay of the respective item-specific memory traces and thus not impede current processing. Albeit speculative, this view would be supported by findings that, compared with HCS, WM maintenance in schizophrenic subjects is more vulnerable to increased delays (46) and PSZ exhibit greater rates of decay in prefrontal brain activation during WM maintenance (47,48). The meta-analysis of Lee and Park (2), however, suggests that delay manipulations do not consistently influence WM performance in PSZ, thus rendering accelerated decay of WM representations an unlikely interpretation. An alternative explanation of the reduced susceptibility to item-specific PI in schizophrenia comes from recent findings that PSZ more efficiently remove deprioritized (no longer relevant) items from WM than HCS (49). As the superior clearance of deprioritized WM contents was found in PSZ to be beneficially related to the maintenance of prioritized WM contents (49), this may suggest even more general differences in (active) control of WM contents in schizophrenia. At first glance, the present data seem to contrast with recent findings that PSZ are impaired in suppressing items that were encoded but subsequently marked as irrelevant (16,50). The applied tasks have, however, a major structural difference in that, in the present recent-probes task, previously relevant WM contents of trial n-1 are deprioritized at the end of the trial and thereafter replaced by the new target items of trial n [a similar principle applies to the study of Hahn et al. (49)], whereas, in the suppress task used by Smith et al. (16) and Eich et al. (50), a subset of an overall set of concurrently encoded target items is devaluated by a cue after a certain delay. Thus, the latter task places high demands on WM selection processes and the categorical allocation of encoded items into relevant and irrelevant ones but does not enforce updating of WM contents. As updating and selection constitute presumably independent processes of WM control (51), present evidence for paradoxically faster clearance of WM contents in schizophrenia is therefore not mutually exclusive with findings of an impaired selection from WM. Given that dopaminergic antagonists are known to impair the stability of WM representations (52,53), it remains to be clarified to which extent a presumable difference in the controlled clearance of WM may be mediated by the patients’ antipsychotic medication (54,55). Already Hemsley (56) proposed “a weakening of the influence of … previous input on current perception“ (p. 182) as a basic cognitive deficit in schizophrenia that may also affect www.sobp.org/journal

8 BIOL PSYCHIATRY 2014;]:]]]–]]] processing of task-relevant information in WM (17). Among others, this assumption is based on observations of abnormal performance of PSZ in latent inhibition (57) and, perhaps more relevant here, in negative priming (18,58). In negative priming, target selection is slowed and less accurate if the target has been a distractor item in the preceding trial. This effect is seen as an indicator of forward inhibitory control in selective attention that prevents recently activated but behaviorally irrelevant information from being immediately re-activated or further processed (59). As PSZ are less affected by negative priming than HCS (58,60), this inhibitory mechanism is believed to be deficient in schizophrenia. However, negative priming is multiply determined (59) and may also be caused by backward memory-based phenomena, such as an automatic retrieval of the most recent encounter with the respective target in the previous trial, including the item’s negative response tag as an irrelevant distractor that interferes with requirements for a positive response in the current trial (59,61,62). Further, given the tasks’ similar temporal structure, the apparent commonality of both effects (i.e., item-specific PI and negative priming) is intriguing in that previously activated information interferes with current attentional processing under healthy conditions but not in schizophrenia. Considerations on cleared activations of memory traces in PI mentioned above may thus also constitute a valid account for the reduced susceptibility to negative priming in schizophrenia (60). Taken together, the present study provides first evidence for a reduced vulnerability of PSZ to item-specific PI that is not due to superior WM control mechanisms but, conversely, a consequence of the patients’ presumably more effective clearance of previous relevant WM traces. Susceptibility to item-specific PI may therefore constitute a promising endophenotype for delineating the cognitive deficits in schizophrenia and their biological foundations. This research was supported in part by a grant of the European Union Seventh Framework Programme for Research (PLASTICISE: Collaborative Project 223524) and the BrainLinks-BrainTools Cluster of Excellence funded by the German Research Foundation (Grant no. EXC 1086). We thank Anna Eras, Tobias Etzold, Sabrina Huber, Isabella Sing, and Tobias Uhl for assistance during data acquisition. All authors report no biomedical financial interests or potential conflicts of interest. Supplementary material cited in this article is available online at http://dx.doi.org/10.1016/j.biopsych.2014.03.012. 1. Goldman-Rakic PS (1994): Working memory dysfunction in schizophrenia. J Neuropsychiatry Clin Neurosci 6:348–357. 2. Lee J, Park S (2005): Working memory impairments in schizophrenia: A meta-analysis. J Abnorm Psychol 114:599–611. 3. Kane MJ, Engle RW (2000): Working-memory capacity, proactive interference, and divided attention: Limits on long-term memory retrieval. J Exp Psychol Learn Mem Cogn 26:336–358. 4. Hasher L, Lustig C, Zacks R (2008): Inhibitory mechanisms and the control of attention. In: Conway A, Jarrold C, Kane M, Miyake A, Towse J, editors. Variation in Working Memory. New York: Oxford University Press. 5. Dempster FN, Corkill AJ (1999): Individual differences in susceptibility to interference and general cognitive ability. Acta Psychol (Amst) 101: 395–416. 6. Lustig C, May CP, Hasher L (2001): Working memory span and the role of proactive interference. J Exp Psychol Gen 130:199–207. 7. Gray JR, Chabris CF, Braver TS (2003): Neural mechanisms of general fluid intelligence. Nat Neurosci 6:316–322.

www.sobp.org/journal

C.P. Kaller et al. 8. Bunting M (2006): Proactive interference and item similarity in working memory. J Exp Psychol Learn Mem Cogn 32:183–196. 9. Burgess GC, Gray JR, Conway AR, Braver TS (2011): Neural mechanisms of interference control underlie the relationship between fluid intelligence and working memory span. J Exp Psychol Gen 140: 674–692. 10. Jonides J, Nee DE (2006): Brain mechanisms of proactive interference in working memory. Neuroscience 139:181–193. 11. Carter CS, Barch DM (2007): Cognitive neuroscience-based approaches to measuring and improving treatment effects on cognition in schizophrenia: The CNTRICS initiative. Schizophr Bull 33:1131–1137. 12. Monsell S (1978): Recency, immediate recognition memory, and reaction time. Cogn Psychol 10:465–501. 13. Barch DM, Berman MG, Engle R, Jones JH, Jonides J, Macdonald A 3rd, et al. (2009): CNTRICS final task selection: Working memory. Schizophr Bull 35:136–152. 14. Barch DM, Moore H, Nee DE, Manoach DS, Luck SJ (2012): CNTRICS imaging biomarkers selection: Working memory. Schizophr Bull 38: 43–52. 15. Mecklinger A, Weber K, Gunter TC, Engle RW (2003): Dissociable brain mechanisms for inhibitory control: Effects of interference content and working memory capacity. Brain Res Cogn Brain Res 18:26–38. 16. Smith EE, Eich TS, Cebenoyan D, Malapani C (2011): Intact and impaired cognitive-control processes in schizophrenia. Schizophr Res 126:132–137. 17. Gray JA, Feldon J, Rawlins J, Hemsley DR, Smith AD (1991): The neuropsychology of schizophrenia. Behav Brain Sci 14:1–84. 18. Hemsley DR (2005): The development of a cognitive model of schizophrenia: Placing it in context. Neurosci Biobehav Rev 29:977–988. 19. Westerhausen R, Kompus K, Hugdahl K (2011): Impaired cognitive inhibition in schizophrenia: A meta-analysis of the Stroop interference effect. Schizophr Res 133:172–181. 20. Friedman NP, Miyake A (2004): The relations among inhibition and interference control functions: A latent-variable analysis. J Exp Psychol Gen 133:101–135. 21. Randolph C, Gold JM, Carpenter CJ, Goldberg TE, Weinberger DR (1992): Release from proactive interference: Determinants of performance and neuropsychological correlates. J Clin Exp Neuropsychol 14: 785–800. 22. Traupmann KL, Berzofsky M, Kesselman M (1976): Encoding of taxonomic word categories by schizophrenics. J Abnorm Psychol 85: 350–355. 23. American Psychiatric Association (2000): Diagnostic and Statistical Manual of Mental Disorders, DSM-IV-TR, 4th ed. Washington, DC: American Psychiatric Association. 24. Andreasen NC (1984): Scale for the Assessment of Positive Symptoms (SAPS). Iowa City: University of Iowa. 25. Andreasen NC (1984): Scale for the Assessment of Negative Symptoms (SANS). Iowa City: University of Iowa. 26. Andreasen NC (1986): Scale for the assessment of thought, language, and communication (TLC). Schizophr Bull 12:473–482. 27. Hamilton M (1960): A rating scale for depression. J Neurol Neurosurg Psychiatry 23:56–62. 28. Young RC, Biggs JT, Ziegler VE, Meyer DA (1978): A rating scale for mania: Reliability, validity and sensitivity. Br J Psychiatry 133:429–435. 29. Ackenheil M, Stotz G, Dietz-Bauer R, Vossen A (1999): Mini International Neuropsychiatric Interview (German Version 5.0.0, DSM-IV). Munich: Psychiatrische Universitätsklinik. 30. Fydrich T, Renneberg B, Schmitz B, Wittchen HU (1997): Strukturiertes Klinisches Interview für DSM-IV. Achse II: Persönlichkeitsstörungen Interviewheft. Göttingen, Germany: Hogrefe. 31. Lehrl S (1999): Mehrfachwahl-Wortschatz-Intelligenztest MWT-B. Balingen, Germany: Spitta Verlag. 32. Sternberg S (1966): High-speed scanning in human memory. Science 153:652–654. 33. Badre D, Wagner AD (2005): Frontal lobe mechanisms that resolve proactive interference. Cereb Cortex 15:2003–2012. 34. Kriegeskorte N, Simmons WK, Bellgowan, Patrick SF, Baker CI (2009): Circular analysis in systems neuroscience: The dangers of double dipping. Nat Neurosci 12:535–540. 35. Leonard CJ, Kaiser ST, Robinson BM, Kappenman ES, Hahn B, Gold JM, Luck SJ (2013): Toward the neural mechanisms of reduced working memory capacity in schizophrenia. Cereb Cortex 23:1582–1592.

C.P. Kaller et al. 36. Jonides J, Smith EE, Marshuetz C, Koeppe RA, Reuter-Lorenz PA (1998): Inhibition in verbal working memory revealed by brain activation. Proc Natl Acad Sci U S A 95:8410–8413. 37. D’Esposito M, Postle BR, Jonides J, Smith EE (1999): The neural substrate and temporal dynamics of interference effects in working memory as revealed by event-related functional MRI. Proc Natl Acad Sci U S A 96:7514–7519. 38. Postle BR, Brush LN, Nick AM (2004): Prefrontal cortex and the mediation of proactive interference in working memory. Cogn Affect Behav Neurosci 4:600–608. 39. Nee DE, Jonides J, Berman MG (2007): Neural mechanisms of proactive interference-resolution. Neuroimage 38:740–751. 40. Bunge SA, Ochsner KN, Desmond JE, Glover GH, Gabrieli JD (2001): Prefrontal regions involved in keeping information in and out of mind. Brain 124:2074–2086. 41. Nelson JK, Reuter-Lorenz PA, Sylvester CC, Jonides J, Smith EE (2003): Dissociable neural mechanisms underlying response-based and familiarity-based conflict in working memory. Proc Natl Acad Sci U S A 100:11171–11175. 42. Oberauer K (2001): Removing irrelevant information from working memory: A cognitive aging study with the modified Sternberg task. J Exp Psychol Learn Mem Cogn 27:948–957. 43. Oberauer K (2005): Binding and inhibition in working memory: Individual and age differences in short-term recognition. J Exp Psychol Gen 134:368–387. 44. Cowan N (1995): Attention and Memory. An Integrated Framework. New York: Oxford University Press. 45. Kane MJ, Conway A, Hambrick DZ, Engle RW (2008): Variation in working memory capacity as variation in executive attention and control. In: Conway A, Jarrold C, Kane M, Miyake A, Towse J, editors. Variation in Working Memory. New York: Oxford University Press, 21–48. 46. Goldberg TE, Patterson KJ, Taqqu Y, Wilder K (1998): Capacity limitations in short-term memory in schizophrenia: Tests of competing hypotheses. Psychol Med 28:665–673. 47. Driesen NR, Leung H, Calhoun VD, Constable RT, Gueorguieva R, Hoffman R, et al. (2008): Impairment of working memory maintenance and response in schizophrenia: Functional magnetic resonance imaging evidence. Biol Psychiatry 64:1026–1034. 48. Anticevic A, Repovs G, Barch DM (2013): Working memory encoding and maintenance deficits in schizophrenia: Neural evidence for activation and deactivation abnormalities. Schizophr Bull 39: 168–178.

BIOL PSYCHIATRY 2014;]:]]]–]]] 9 49. Hahn B, Hollingworth A, Robinson BM, Kaiser ST, Leonard CJ, Beck VM, et al. (2012): Control of working memory content in schizophrenia. Schizophr Res 134:70–75. 50. Eich TS, Nee DE, Insel C, Malapani C, Smith EE (2013): Neural correlates of impaired cognitive control over working memory in schizophrenia [published online ahead of print October 10]. Biol Psychiatry. http://dx. doi.org/10.1016/j.biopsych.2013.09.032. 51. Bledowski C, Kaiser J, Rahm B (2010): Basic operations in working memory: Contributions from functional imaging studies. Behav Brain Res 214:172–179. 52. Sawaguchi T, Goldman-Rakic PS (1991): D1 dopamine receptors in prefrontal cortex: Involvement in working memory. Science 251: 947–950. 53. Sawaguchi T, Goldman-Rakic PS (1994): The role of D1-dopamine receptor in working memory: Local injections of dopamine antagonists into the prefrontal cortex of rhesus monkeys performing an oculomotor delayed-response task. J Neurophysiol 71:515–528. 54. Barch DM (2006): What can research on schizophrenia tell us about the cognitive neuroscience of working memory? Neuroscience 139:73–84. 55. Rolls ET, Loh M, Deco G, Winterer G (2008): Computational models of schizophrenia and dopamine modulation in the prefrontal cortex. Nat Rev Neurosci 9:696–709. 56. Hemsley DR (1987): An experimental psychological model for schizophrenia. In: Hafner H, Gattaz WF, Janzavik W, editors. Search for the Causes of Schizophrenia. Heidelberg, Germany: Springer, 179–188. 57. Baruch I, Hemsley DR, Gray JA (1988): Differential performance of acute and chronic schizophrenics in a latent inhibition task. J Nerv Ment Dis 176:598–606. 58. Beech A, Powell T, McWilliam J, Claridge G (1989): Evidence of reduced ‘cognitive inhibition’ in schizophrenia. Br J Clin Psychol 28:109–116. 59. May CP, Kane MJ, Hasher L (1995): Determinants of negative priming. Psychol Bull 118:35–54. 60. Peters ER, Pickering AD, Kent A, Glasper A, Irani M, David AS, et al. (2000): The relationship between cognitive inhibition and psychotic symptoms. J Abnorm Psychol 109:386–395. 61. Neill WT, Valdes LA, Terry KM, Gorfein DS (1992): Persistence of negative priming: II. Evidence for episodic trace retrieval. J Exp Psychol Learn Mem Cogn 18:993–1000. 62. Fox E (1995): Negative priming from ignored distractors in visual selection: A review. Psychol Bull Rev 2:145–173. 63. Snodgrass JG, Vanderwart M (1980): A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. J Exp Psychol Hum Learn 6:174–215.

www.sobp.org/journal