Introduction
Memory dysfunction in verbal and non-verbal domains constitutes a substantial part of the cognitive impairment observed in patients with schizophrenia (Aleman, Hijman, de Haan, & Kahn, Reference Aleman, Hijman, de Haan and Kahn1999; Heinrichs & Zakzanis, Reference Heinrichs and Zakzanis1998). The deficit appears to pertain mostly to the processes of encoding the information (Cirillo & Seidman, Reference Cirillo and Seidman2003; Mesholam-Gately, Giuliano, Goff, Faraone, & Seidman, Reference Mesholam-Gately, Giuliano, Goff, Faraone and Seidman2009). Remediation of this deficit is crucial, since memory efficiency is a strong determinant of functional outcome in this population (Green, Reference Green1996). To develop more appropriate and effective cognitive and pharmacological remediation techniques, it is important to determine the causes of the impaired performance observed in patients.
Memory deficit might be primary to the disorder. Alternatively, it might result from dysfunction in other underlying processes. Psychomotor slowing is a consistent and perhaps core feature of schizophrenia patients (Dickinson, Ramsey, & Gold, Reference Dickinson, Ramsey and Gold2007; Knowles, David, & Reichenberg, Reference Knowles, David and Reichenberg2010; Morrens, Hulstijn, & Sabbe, Reference Morrens, Hulstijn and Sabbe2007). Decreased processing speed has been identified by the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) committee as one of the fundamental cognitive deficits in this population (Nuechterlein et al., Reference Nuechterlein, Barch, Gold, Goldberg, Green and Heaton2004). This fundamental deficit has been implicated in the memory impairment observed in these patients. Notably, decreased processing speed has been linked to poor verbal memory efficiency in several schizophrenia samples (Holthausen et al., Reference Holthausen, Wiersma, Sitskoorn, Dingemans, Schene and van den Bosch2003; Leeson et al., Reference Leeson, Barnes, Harrison, Matheson, Harrison, Mutsatsa and Joyce2010; Morrens, Hulstijn, Van Hecke, Peuskens, & Sabbe, Reference Morrens, Hulstijn, Van Hecke, Peuskens and Sabbe2006; Ojeda, Peña, Sánchez, Elizagárate, & Escurra, Reference Ojeda, Peña, Sánchez, Elizagárate and Escurra2008; Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias, & Vazquez-Barquero, Reference Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias and Vazquez-Barquero2007; Sanfilipo et al., Reference Sanfilipo, Lafargue, Rusinek, Arena, Loneragan, Lautin and Wolkin2002). More specifically, our group observed that slowing affects the encoding processes of verbal memory (Brébion et al., Reference Brébion, Smith, Gorman, Malaspina, Sharif and Amador2000; Brébion, David, Bressan, & Pilowsky, Reference Brébion, David, Bressan and Pilowsky2007). A link between processing speed and visual memory has been established too (Brébion, Bressan, David, & Pilowsky, Reference Brébion, Bressan, David and Pilowky2009; Holthausen et al., Reference Holthausen, Wiersma, Sitskoorn, Dingemans, Schene and van den Bosch2003; Sanfilipo et al., Reference Sanfilipo, Lafargue, Rusinek, Arena, Loneragan, Lautin and Wolkin2002). Controlling for a processing speed measure appears to either eliminate or reduce significantly the deficit observed relative to a healthy control group in both verbal (Brébion, Gorman, Malaspina, Sharif, & Amador, Reference Brébion, Gorman, Malaspina, Sharif and Amador2001; Brébion, David, Bressan, & Pilowsky, Reference Brébion, David, Bressan and Pilowsky2006; Brébion et al., Reference Brébion, David, Bressan and Pilowsky2007; Holthausen et al., Reference Holthausen, Wiersma, Sitskoorn, Dingemans, Schene and van den Bosch2003; Leeson et al., Reference Leeson, Barnes, Harrison, Matheson, Harrison, Mutsatsa and Joyce2010; Rodriguez-Sanchez et al., Reference Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias and Vazquez-Barquero2007) and visual (Brébion et al., Reference Brébion, Bressan, David and Pilowky2009; Holthausen et al., Reference Holthausen, Wiersma, Sitskoorn, Dingemans, Schene and van den Bosch2003) memory. Processing speed has also been involved in verbal fluency (Kremen, Seidman, Faraone, & Tsuang, Reference Kremen, Seidman, Faraone and Tsuang2003; Ojeda et al., Reference Ojeda, Peña, Sánchez, Elizagárate and Escurra2008; Rodriguez-Sanchez et al., Reference Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias and Vazquez-Barquero2007; Sanfilipo et al., Reference Sanfilipo, Lafargue, Rusinek, Arena, Loneragan, Lautin and Wolkin2002; van Beilen et al., Reference van Beilen, Pijnenborg, van Zomeren, van den Bosch, Withaar and Bouma2004) and other executive functions (Leeson et al., Reference Leeson, Barnes, Harrison, Matheson, Harrison, Mutsatsa and Joyce2010; Morrens et al., Reference Morrens, Hulstijn, Van Hecke, Peuskens and Sabbe2006; Rodriguez-Sanchez et al., Reference Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias and Vazquez-Barquero2007) in patients. It has been proposed as a crucial factor of functional outcome (Sánchez et al., Reference Sánchez, Ojeda, Peña, Elizagárate, Yoller, Gutiérrez and Escurra2009).
Another potential mechanism of cognitive impairment is decreased working memory span. Working memory has been defined as a capacity-limited system which enables the simultaneous short-term storage and processing of the information (Baddeley, Reference Baddeley1986). It involves a central executive component, in charge of the manipulation of the information that is being temporally maintained. Meta-analyses have revealed working memory impairment in schizophrenia (Forbes, Carrick, McIntosh, & Lawrie, Reference Forbes, Carrick, McIntosh and Lawrie2009; Lee & Park, Reference Lee and Park2005). The processes of working memory consolidation notably appear to be abnormal (Fuller et al., Reference Fuller, Luck, Braun, Robinson, McMahon and Gold2009). Decreased working memory span has been proposed as a core deficit that underlies dysfunction in a broad range of neuropsychological domains in this population (Ojeda et al., Reference Ojeda, Sánchez, Peña, Elizagárate, Yoller, Larumbe and Ezcurra2010; Silver, Feldman, Bilker, & Gur, Reference Silver, Feldman, Bilker and Gur2003; Silver & Goodman, Reference Silver and Goodman2008). Stone, Gabrieli, Stebbins, and Sullivan (Reference Stone, Gabrieli, Stebbins and Sullivan1998) observed that decreased working memory span accounted for the patients’ deficit in the free-recall of semantically organizable lists of words, and other verbal memory tasks that required self-initiated strategy. By contrast, it did not contribute significantly to deficits in word recognition, a task that does not rely on such strategy and, therefore, appears less effortful.
The purpose of this study was to investigate the relative role of processing speed and working memory on verbal and visual memory efficiency in schizophrenia. The role of processing speed has already been established by our group and others. To extend our work on the cognitive underpinnings of memory impairment in schizophrenia, we sought to test the hypothesis that working memory is additionally involved in memory efficiency, especially insofar as effortful encoding processes are involved. The distinction between superficial and deep, effortful encoding processes is particularly meaningful, since memory efficiency strongly relies upon the depth of encoding (Craik, Reference Craik2002). One common way of investigating depth of processing in verbal memory is to contrast serial versus semantic encoding in lists of words that contain a latent semantic organization. Serial clustering reflects the superficial strategy of learning the list of words by rote rehearsal. Semantic clustering reflects a deeper processing of the semantic properties of the words. In a broad study of verbal and visual memory in schizophrenia, we used lists of words with semantic organization, and examined the serial and semantic strategies implemented to learn them. To extend the investigation of superficial versus deep encoding, we also prepared lists in which high- and low-frequency words were intermixed. While the processing of high-frequency words is relatively automatic, low-frequency words, which are not familiar to the reader, require more effort to process, draw more attention, and consume more resources (Malmberg & Nelson, Reference Malmberg and Nelson2003). Finally, in a visual recognition task, we presented mixed colored and black-and-white pictures. A few studies have revealed an advantage for colored pictures, attributed to their richer characteristics and the fact that they require more attentional processes at encoding (Wichmann, Sharpe, & Gegenfurtner, Reference Wichmann, Sharpe and Gegenfurtner2002). The colored pictures are presumably more deeply processed than the other ones. To increase this differential processing we displayed simultaneously four mixed pictures of both types, and the subjects freely allocated their attention and study time to the stimuli during the exposure duration. The colored pictures, more distinctive and salient, drew more attention to the detriment of the other pictures, which were, therefore, processed more superficially. Indeed, distinctiveness of the items is a key feature of depth of encoding within the level-of-processing framework (Hunt & McDaniel, Reference Hunt and McDaniel1993).
We have already reported in all or part of the same patient and healthy control samples the role of processing speed on the recall of non-organizable and semantically organizable lists of words (41 of the patients, 41 of the controls, Brébion et al., Reference Brébion, David, Bressan and Pilowsky2006), recall and recognition of high- and low-frequency words (48 of the patients, 41 of the controls, Brébion et al., Reference Brébion, David, Bressan and Pilowsky2007), and picture recognition (all 49 patients and 43 controls, Brébion et al., Reference Brébion, Bressan, David and Pilowky2009). The role of processing speed on the serial and semantic encoding strategies that reflect depth of verbal processing was not investigated in this sample. The role of working memory on any memory process was not sought either. In the current study, we conducted regression analyses on the various superficial and deep memory measures derived from the task, using processing speed and working memory span simultaneously as predictors. Three tests of processing speed were used in the battery, so that in previous analyses we were able to investigate the specific role of cognitive and motor speed on diverse aspects of verbal memory. In the current study, scores from the three tests were combined into a single processing speed measure. We expected processing speed to contribute to the serial and semantic strategy indices in patients, as observed in a previous independent sample (Brébion et al., Reference Brébion, Smith, Gorman, Malaspina, Sharif and Amador2000). We also expected it to contribute to the recall and recognition indices in verbal and visual memory, corroborating the previous analyses that did not include working memory span. These findings would suggest that processing speed has an effect of its own on various aspects of memory, insofar as the previously observed associations were not confounded by potential overlap between processing speed and working memory. Following Stone et al. (Reference Stone, Gabrieli, Stebbins and Sullivan1998), we expected working memory span to make an additional significant contribution to the measures reflecting deep memory encoding, in both verbal and visual memory. By contrast, working memory span was not expected to be associated with any measure of superficial encoding. Whether a similar pattern of associations with working memory span is observed in healthy controls was investigated. Finally, we conducted regression analyses in patients and healthy controls altogether, using processing speed and working memory span alternately as predictors, in addition to diagnosis. This enabled us to determine the extent to which decrement in each of these functions accounted for the deep encoding deficit in patients.
Method
Subjects
Forty-nine patients with schizophrenia (8 in-patients) were recruited at the Maudsley Hospital, London (32 males, 17 females; 26 Caucasian, 19 black, 4 Indian/Asiatics; age: mean = 34.7; SD = 7.8; number of years of education: mean = 12.7; SD = 2.6; National Adult Reading Test (NART) score: mean = 100.9; SD = 13.9). The diagnosis was made on the basis of Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria by two experienced psychiatrists who used clinical interview, patients’ history and chart review, and reached a consensus. The patients were primarily suffering from chronic schizophrenia, with disease duration of over 12 months. Exclusion criteria for the study were any evidence of alcohol or drug abuse (DSM-IV criteria), organic mental illness or mental impairment, history of brain injury, and current severe physical illness. Symptoms ratings were available for 41 of the patients (Scale for the Assessment of Positive Symptoms: mean = 15.8; SD = 17.8; Scale for the Assessment of Negative Symptoms: mean = 20.3; SD = 14.6). All patients except three were on daily antipsychotic medication (olanzapine, quetiapine, amisulpiride, clozapine, risperidone, haloperidol, zuclopenthixol).
Forty-three healthy comparison subjects were recruited by local announcements and among hospital staff, and were screened to rule out any current or recent psychiatric history (27 males, 16 females; 28 Caucasian, 13 black, 2 Indian/Asiatics; age: mean = 35.0; SD = 9.6; number of years of education: mean = 12.9; SD = 2.8; NART score: mean = 105.7; SD = 11.9). The two groups were not significantly different with respect to age, gender distribution, ethnicity, education level, and premorbid IQ assessed by NART score. Ethical approval for the study was obtained from the local hospital Research Ethics Committee. After a full explanation of the study, subjects provided written informed consent to participate.
Material and Procedure
Verbal memory
Serial versus semantic encoding
Four lists of 16 concrete mono- or bisyllabic words organizable into four semantic categories (Battig & Montague, Reference Battig and Montague1969) were prepared. They were equivalent in average number of syllables. Two of them were made up of four typical instances within each category, whereas the other two were made up of four atypical instances. The words were distributed randomly within the lists, so that the semantic organization was not obvious. The subjects were presented with one typical and one atypical semantically organizable list. They were not informed of the latent semantic organization. The choice of the target lists, as well as their order of presentation, was counterbalanced among subjects.
The lists of words were printed out. One organizable list was first presented. The subjects were informed that they had 45 seconds to learn the words, and were required to read the list aloud at least once. Immediately after learning, they were provided with a blank sheet and required to write down as many words as they could remember in any order. No time limit was imposed for free recall. Then, the second organizable list was administered following the same procedure.
The total number of words recalled for the two organizable lists was tallied. A serial clustering score was computed from the recall of each list (proportion of words recalled in their order of presentation, out of the number of recalled words). It indexes the propensity to learn the list by rote rehearsal, which reflects superficial encoding. The efficiency of this strategy was assessed by the number of words recalled in sequence for each list. A semantic clustering score was also computed from the recall of each list (number of words following a word of the same semantic category, out of the number of possible semantic associations in the subject's output). This measure reflects the propensity to carry out deep encoding through the use of semantic organization. The efficiency of this strategy was assessed in each list by the number of categories for which semantic organization was observed, and by the number of recalled words within each recalled category. These serial and semantic encoding indices were averaged across the two organizable lists.
Memory for high- versus low-frequency words
Four non-organizable lists of 16 concrete mono- or bisyllabic words were prepared. Each list was made up of eight high-frequency words and eight low-frequency words (Kucera & Francis, Reference Kucera and Francis1967) in random order. For the high-frequency words (e.g., FLOWER, BOAT, CHURCH) the occurrence was more than 42 per million (mean for the four lists: 108.6); for the low-frequency words (e.g., HAMMOCK, EASEL, SHIELD) the occurrence was less than 10 per million (mean for the four lists: 6.6). The four lists were equivalent in average number of syllables. The mean frequency of the eight high-frequency words was equivalent in the four lists, as was the mean frequency of the eight low-frequency words. One of these lists was randomly chosen as a target list to be presented for immediate free recall and recognition. A recognition sheet consisting of a list of 32 words was prepared. It included in random order the 16 words used in the target list and the 16 words of another list, used as distractors. Each of the four lists was alternately selected to be used as a target or as a distractor list, in a counterbalanced way.
The target list was presented for 45 seconds, then immediate free recall was requested. The subjects were then given the recognition sheet and instructed to circle all the words they could recognize from the list.
The number of high- and low-frequency target words correctly recalled was tallied. In the recognition task, the number of correctly recognized high- and low-frequency target words, as well as the number of erroneously recognized high- and low-frequency distractor words, was tallied. These measures were combined to compute, for each type of word, a recognition index Pr reflecting the accuracy in discriminating target words from distractors (rate of correct recognitions of target words minus rate of false recognitions of non-target words; Corwin, Reference Corwin1994). The encoding of the low-frequency words was assumed to require deeper processing than that of the high-frequency words.
Visual memory
The visual stimuli were 32 black-and-white (b/w) and colored pictures from art galleries, all representational, selected among those that were not famous. Sixteen pictures (8 b/w and 8 colored) were presented as targets, and the remaining 16 (8 b/w and 8 colored) were used as distractors. The use of each picture as target or distractor was counterbalanced among subjects, as was the presentation of each picture in its black-and-white or colored version.
The stimuli were laid on the table in groups of four mixed b/w and colored pictures, and displayed over 20 seconds. The 16 pictures were thus presented. After a delay of approximately 5 minutes, the 16 target pictures mixed with the 16 distractors were presented one by one in random order. The subjects were required to indicate, for each of them, whether it had been presented during the acquisition phase, or was new.
The number of recognized target pictures, as well as the number of false recognitions of new pictures, was tallied for b/w and for colored pictures. The recognition index Pr reflecting accuracy in discriminating target pictures from distractors was computed for each type of picture. The encoding of the colored pictures was assumed to require deeper processing than that of the black-and-white pictures.
Working memory
The backward digit span from the Wechsler Adult Intelligence Scale-Revised (WAIS-R) was administered. Increasingly long series of digits had to be repeated in reverse order. The total number of correct trials was tallied.
Processing speed
Three processing speed tests were used: The Digit Symbol Substitution Test from the WAIS-R, the Trail Making Test-A, and the color naming neutral condition of the Stroop test. The measures derived from these tests were z-transformed using the mean and standard deviation from the control group. The sign of the response time from the latter two tests was inversed, and the three resulting measures were averaged to create a global processing speed measure.
Results
The scores obtained in the patient and in the control groups for all the studied cognitive variables are presented in Table 1.
Table 1 Comparison between schizophrenia patients and healthy controls on the cognitive measures
Note. Means, standard deviations, and t-test comparisons.
Role of Processing Speed and Working Memory Span on Serial versus Semantic Clustering
Regression analyses were conducted in each group. The dependent variables were the two indices of serial clustering, reflecting superficial encoding; the three indices of semantic clustering, reflecting deep encoding; and the total number of words recalled in the two organizable lists, which was assumed to rely heavily on deep encoding. The global processing speed measure and the backward digit span were entered simultaneously as predictors. Age was entered as well to control for its effect on the memory measures. Results are presented in Table 2.
Table 2 Regression analyses in the 49 schizophrenia patients and in the 43 healthy controls
Note. Associations of processing speed and working memory span with serial and semantic clustering indices (β coefficient). p < .10, *p < .05, **p < .01, ***p < .001, ****p < .0001.
In patients, both processing speed and working memory span were highly significant predictors of the number of words recalled in the organizable lists. An examination of the strategy indices reveals that processing speed tended to contribute to the serial clustering score, and made a significant contribution to the number of words recalled in sequence. Similarly, it tended to contribute to the semantic clustering score, and made a significant contribution to the two indices reflecting efficiency of semantic encoding. Working memory span was not a predictor of either of the serial strategy indices. If anything, the association went in the opposite direction. By contrast, it was a predictor of the three indices of semantic clustering, as expected, although the association with the semantic clustering score was only observed at a trend level of significance.
In controls, the effect of working memory was fairly similar to that observed in patients, in that working memory span contributed to semantic, but not serial, clustering strategy. The effect of processing speed appears weaker and less extensive than in patients.
Role of Processing Speed and Working Memory Span on Memory for High- versus Low-Frequency Words
Regression analyses were conducted in each group on the number of recalled words and recognition index Pr for the high- and for the low-frequency words. The processing speed measure, working memory span, and age were entered simultaneously as predictors. Results are presented in Table 3. As expected, in the patient group, working memory span contributed significantly to the recall and recognition of the low-frequency words, assumed to rely on deep encoding, but not to either recall or recognition of the high-frequency words. A similar pattern of associations with working memory span was observed in the control group. Processing speed contributed significantly to the recall of both types of word in patients, but only to that of the high-frequency words in healthy controls.
Table 3 Regression analyses in the 49 schizophrenia patients and in the 43 healthy controls
Note. Associations of processing speed and working memory span with recall and recognition of the high- and low-frequency words (β coefficient). p < .10, *p < .05, **p < .01.
Role of Processing Speed and Working Memory Span on the Recognition of Black-and-White versus Colored Pictures
Regression analyses were conducted on the recognition indices Pr derived from the visual memory task. Results are presented in Table 4. In both groups, working memory span contributed significantly to the recognition of the colored pictures, assumed to request deeper encoding than the b/w pictures. Processing speed contributed significantly to the recognition of both types of picture in patients, but only to that of the b/w pictures in controls.
Table 4 Regression analyses in the 49 schizophrenia patients and in the 43 healthy controls
Note. Associations of processing speed and working memory span with recognition of the black-and-white and colored pictures (β coefficient). p < .10, *p < .05, **p < .01.
Contribution of Decreased Processing Speed and Decreased Working Memory Span to Memory Deficit in Patients
Regression analyses involving all the participants were conducted on the memory measures relying on deep encoding processes. The purpose of these analyses was to determine the extent to which the memory deficit in patients is reduced by controlling either processing speed or working memory span. Diagnosis was first entered as the sole predictor. Then diagnosis and the processing speed measure were entered simultaneously. Finally, diagnosis and working memory span were entered simultaneously. Results are presented in Table 5.
Table 5 Regression analyses on the memory measures relying on deep encoding
Note. Effect of diagnosis (β coefficient and p value) when diagnosis was the sole predictor; then when diagnosis and processing speed measure were entered simultaneously as predictors; then when diagnosis and working memory span were entered simultaneously as predictors.
When diagnosis was entered alone, it had a significant effect on all variables. This indicates that all these measures were significantly impaired in patients. When the processing speed measure was added to the regression analyses, the effect of diagnosis was no longer significant for any of these measures, suggesting that decreased processing speed accounts for the observed deficits in patients. When working memory span, instead of processing speed, was added to diagnosis, the effect of diagnosis was reduced. However, this effect remained significant for all measures, except the semantic clustering score for which only a trend was observed. The fact that the effect of diagnosis remains significant after adding working memory span to the regression analyses suggests that patients would still demonstrate impairment in the studied memory measures if their working memory span were equivalent to that of the healthy controls.
Discussion
In this study, we have used different ways of varying the depth of encoding, in memory tasks that involved verbal as well as visual stimuli. The associations of two potential cognitive underpinnings, processing speed and working memory span, with the superficial and deep encoding processes were investigated simultaneously. While the role of processing speed on various aspects of memory efficiency had previously been reported in this patient sample, we wished to test the hypothesis that working memory has an additional and more specific effect.
Regression analyses show that in patients with schizophrenia, processing speed had a pervasive effect on the superficial and deep encoding processes of verbal and visual memory. In the learning of lists of words, processing speed contributed to the efficiency of both rote rehearsal and semantic organization, corroborating the findings from a previous independent sample (Brébion et al., Reference Brébion, Smith, Gorman, Malaspina, Sharif and Amador2000). Processing speed also contributed to the recall of the low-frequency words, presumably requiring effortful encoding, as well as to that of the more superficially encoded high-frequency words. Finally, it had a role in the recognition of both black-and-white and colored pictures. The effect of processing speed was significant in the healthy controls too, albeit more restricted than in the patient group. Patients demonstrated significant impairment on all the memory measures reflecting deep encoding. However, controlling for processing speed reduced to non-significance the effect of diagnosis on all these measures. Significant reduction of verbal and visual memory deficit after controlling for processing speed was similarly observed in previous research by our group and others (Brébion et al., Reference Brébion, Gorman, Malaspina, Sharif and Amador2001; Holthausen et al., Reference Holthausen, Wiersma, Sitskoorn, Dingemans, Schene and van den Bosch2003; Leeson et al., Reference Leeson, Barnes, Harrison, Matheson, Harrison, Mutsatsa and Joyce2010; Rodriguez-Sanchez et al., Reference Rodriguez-Sanchez, Crespo-Facorro, Gonzalez-Blanch, Perez-Iglesias and Vazquez-Barquero2007). All these findings suggest that decreased processing speed is a crucial factor in verbal and visual memory deficiency in patients.
As expected, working memory was involved in the deep encoding memory processes. Its capacity appears to be a limiting factor in the ability to carry out effortful operations, irrespective of diagnosis status. When the extensive and possibly overlapping effect of processing speed was controlled, working memory span made a significant contribution to the efficiency of semantic organization in the learning of lists of words, as well as to the recall and recognition of the low-frequency words. It was also involved in the recognition of the colored pictures. These associations were similarly observed in the patient and in the healthy control groups. In neither group was working memory span associated with the superficial processes of rote rehearsal or of learning the high-frequency words, or with the encoding of the black-and-white pictures.
Thus, our hypothesis of a significant role of working memory in effortful memory processes, additional to that of processing speed, was confirmed. It is also worth noting that verbal working memory span was involved in both verbal and visual memory efficiency. The backward digit span might not heavily tax the central executive component of working memory. However, it is a sensitive index of working memory impairment in schizophrenia patients (Conklin, Curtis, Katsanis, & Iacono, Reference Conklin, Curtis, Katsanis and Iacono2000), and proved effective in demonstrating a significant contribution to the deep memory measures in our data. Our results are compatible with those of Stone et al. (Reference Stone, Gabrieli, Stebbins and Sullivan1998), who demonstrated the role of working memory in strategic verbal memory. We have extended these findings by showing that working memory is not only related to strategic memory—semantic organization in our task—but to a wider range of deep encoding processes. In addition, we have established that the association also pertains to visual memory. It is likely that the involvement of working memory increases with the degree of effort required to perform the task. Whether the differential association of working memory span with deep versus more superficial processes is observed in cognitive functions others than memory should be investigated in future research.
Regression analyses involving all the participants indicate nonetheless that, although working memory contributes to the efficiency of deep encoding processes, it does not account for the observed deficit in patients. A stronger reduction of the deficit after adjusting working memory span might be revealed if a more demanding working memory task were used (Stone et al., Reference Stone, Gabrieli, Stebbins and Sullivan1998). In addition, the backward digit span has a restricted range, with little variance. A composite working memory measure derived from several working memory tasks might be more appropriate in future studies.
In schizophrenia patients and healthy controls, both processing speed and working memory span are involved in memory efficiency measures relying on deep encoding. A role of working memory and processing speed decrement in memory impairment has similarly been observed in elderly participants (Salthouse, Reference Salthouse1992) as well as in depressed patients (Nebes et al., Reference Nebes, Butters, Mulsant, Pollock, Zmuda, Houck and Reynolds2000). However, the deficit in our schizophrenia patients is accounted for by decreased processing speed and not by decreased working memory span. Our conclusions are limited by the lack of explicit manipulation of depth of encoding in the memory tasks, and by the use of a single working memory task with a relatively undemanding manipulation component. In addition, our data merely reveal associations among variables. We can only posit that the observed decrements in processing speed and working memory span have a deleterious effect on memory efficiency in patients. Nonetheless, these findings have implications for the understanding of memory deficit in schizophrenia and its cognitive remediation. They suggest that patients might enhance their memory efficiency if they were allowed more time for the encoding of the stimuli. In addition, cognitive rehabilitation programs focusing on working memory training should be encouraged (Levaux et al., Reference Levaux, Vezzaro, Larøi, Offerlin-Meyer, Danion and Van der Linden2009), since such training might improve daily life functioning in situations in which demanding cognitive operations are required.
Acknowledgments
We are sad to report the untimely death of Professor Pilowsky. The first author was funded by grants from the Leverhulme Trust, the Alexander Gralnick Grant for Research on Schizophrenia, American Psychological Foundation, the Bial Foundation, the British Academy, NARSAD, and the Wodecroft Foundation. The authors declare no conflict of interest.
