Attention is often an automatic process related to the monitoring and awareness of the world around us [33] . Automatic versus effortful and controlled attention is also an integral concept of a theoretical framework called “two-process theory of attention”, which makes a distinction between a focused attention mode when identifying targets and a diffuse attention mode when scanning non-targets in a sustained attention task. In the two-process theory of attention, rapid automatic attention scanning goes on in parallel with other processing, and generates interrupts when a target is identified, while controlled processing is a slower and exclusively serial short term memory process. However, when a person is in a focused state of attention, the person is less receptive to these automatic interrupts [34] . The automatic and controlled processes are assumed separate processes that may occur simultaneously, and depend on interlinked but separate brain processes. Attention and short term memory capacity in depression has been found to primarily be associated with controlled, focused or effortful processing [35] , similarly short-term memory retrieval in effortful attentional processing is also impaired in depression [8] [36] [37] while automatic processing only are affected under severe conditions [38] .

Sustained attention or vigilance requires a combination of both automatic and controlled processing depending upon the task [39] . Decline in ability to sustain attention may reflect dependency upon work-load intensity factors and results may vary between tasks. Sustained attention is important both for simple routine tasks where losing attention could lead to mistakes, as well as more difficult specialized tasks which require a higher level of concentration. One frequently used test is the Trail-making test A (TMT-A), which comes in several versions [31] [40] [41] , e.g. from the Halsted-Reitan battery [42] . The TMT-A requires the participant to draw a trail between numbers on a sheet of paper and requires several additional cognitive abilities, such as information processing, sequencing ability, and motor speed. Thus, it is important that these other cognitive factors must be controlled for in a group analysis. There are several tailor made tests available that measures sustained attention, one is the Continuous Performance Test (CPT, [43] ) which is available in both auditory and visually presented versions. The CPT requires response to a non-frequent target letter every time it appears in a stream of letters, but no response to frequent non-target letters. Another similar test is the Sustained Attention to Response Task (SART test; [27] [44] ). In the SART test, the subjects are required to restrain from responding to a fraction of the targets, while they are required to respond to the rest of the targets, and accuracy and response times are measured. This test show poorer sustained attention even in moderately depressed men [27] . However, there is still not conclusive evidence to claim that sustained attention decline as a function of aging [45] or as a function of depression [2] .

It has long been known that arousal is an important component in order to facilitate sustained attention to stimuli [46] , and arousal depend upon subcortical brain regions including the amygdala that also show deviating activation results for depressed versus normal subjects in brain imaging studies. One prefrontal region, the cholinergic system, releases acetylcholine under normal operation. It has been shown that blocking release of acetylcholine impairs the ability of sustained attention [47] . In addition, rise in acetylcholine level has been found to invoke more severe depression, implicating normal cholinergic release as important in depression and other affective disorders [48] .

The typical example in order to describe the divided attention situation is a cocktail party [49] , where several conversations may go on in parallel within hearing distance, and it is possible to tune in to one particular conversation while registering bits and pieces from other conversations. Attention may be switched if something interesting comes up in any conversation within hearing distance. In the visual mode, divided attention refers to changing viewpoint, e.g., focusing towards movement in the peripheral field of view when driving a car. In the dichotic listening test, consonant-vowel syllables are presented in headphones to both ears, and the participant are asked either to attend to stimuli from a particular ear, or to attend to stimuli without a reference to lateralization of the input. A right ear preference is found in normal adults, however the right ear advantage (REA) is altered in several patient groups, including in schizophrenia where lack of REA is interacting with the occurrence of auditory hallucinations [50] . A similar pattern of missing asymmetry in dichotic listening tests has been reported in major depression, particularly for patients with withdrawal-retardation symptomatology [51] . It is possible that the combination of the affective and attentional reductions due to depression contributed to lower accuracy on the task in the depressed group as compared to the control group, and this may be related to a lack in top-down control of the cognitive resources, including the resources required for maintaining attention to tasks or stimuli.

Divided visual field tasks have shown reduced visual divided attention in MCI patients and Alzheimer patients [52] , and a similar pattern of reduced divided attention may be predicted in depression, since depressed patients may be lacking the ability of focusing visual attention towards the appropriate spatial position of the target, based upon the assumption of a similar top-down control deficit in depression for the visual mode, as was suggested for the auditory mode. Hammar et al. [53] found effortful processing to be reduced in two testing sessions despite improved scores on a depression rating scale (HAM-D; [54] ) between sessions when performing a visual search paradigm. All patients had a history of recurrent depression. In addition, Majer et al. [55] found that performance on a dual auditory / visual divided attention task (part of “Test batterie zur Aufmerk-samkeitsprüfung”, “TAP”) was reduced in depressed patients, suggesting that depression causes deficits in divided attentional performance regardless of perceptual mode. They found a direct correlation between divided attention task performance and depression. The study controlled for medication, depression severity (depression rating score) and depression duration.

If a person is preoccupied with depressive thoughts, the threshold for registering a new event is elevated. This leads to interrupts in the processes leading up to memory formations, with shallow encoding of events, and lesser formation of enduring memories [21] [57] . Since processing speed is one function that naturally is showed in aging, memory deficits as a consequence of depression tend to be even more severe in elderly, due to even shallower encoding of new events [57] [58] . Episodic memory impairment in depression is associated with shallow encoding and poorer effortful control of encoded information [59] . Baune et al. [22] performed a principal components analysis in order to extract the memory components from several tests on elderly, and they did not report significant effects on episodic memory due to depression when other effects were controlled for. Thus, memory problems in depression are at least partly, if not exclusively based on other underlying cognitive functions.

The ability to “remember” implies a controlled recollection or recognition of the encoding event or study episode, while “know” reflect a more automatic and effortless processing. “Remember” and “know” responses reflect two recollection mechanisms which have been found to be functionally independent in that remembering depend more on perceptual factors and knowing depend more on conceptual factors even though there are overlaps [60] [61] . Experiments with the remember-know paradigm show reduced ability of making “remember” responses in normal aging while “know” responses are unaffected [62] . Yonelinas et al. [62] found an association between less elaborative encoding and reduced remembering, and they suggested that older participants were less able to use the appropriate contextual retrieval processes. A similar pattern was found in major depression [63] , suggesting again that the effect of impaired recognition memory should be even stronger in late life depression, due to impairment in recollection processing rather than familiarity processing. The underlying factors may be shallow encoding due to slowed information processing, but it is also possible that independent factors that affect strategic memory planning or other executive planning are affected in depression. Since independent aging [62] and depression studies [63] show similar pattern of results using the remember/know paradigm, it may be assumed that aging and depression affect the same underlying factors.

Updating of short term memory is an information processing function that relies on sufficient processing speed. It is often treated as an executive sub-function or as an attentional control function, since it involves an active strategy in order to focus attention, refreshing, and monitoring what is contained in the limited working memory [64] . Updating of working memory has been studied with n-back go/no-go tasks where the participant decides whether a letter is the same as the one presented n letter earlier in a list (e.g., [24] ), or with serial list recall tasks with verbal or spatial stimuli [12] . Updating of short term memory has been found to be impaired in both depression and in normal aging, and thus updating of short term memory is particularly impaired in late life depression [24] . However, since updating of short-term memory depends on processing speed and attentional control, impairment on short-term memory task are likely to reflect impairment in these sub-functions.

It has been reported specific effects on verbal memory in late life depression, where recalling list learning items is affected while recalling storytelling is unaffected [65] . Lamar et al. [65] compared brain area volumes in late life depression with matched healthy controls, and found a correlation between volume sizes of prefrontal and cingulate sub-regions with poor performance on the list-learning task (California Verbal Learning test, CVLT) in the late life depression group. These brain areas are important for all cognitive functioning. The study show a dissociation between list-learning performance and recollection of storytelling analogous with the dissociation between remembered and known responses in the remember/know paradigm. One common factor is that more structured and effortful encoding paradigms influence subsequent recollection more than it influence automatic or less structured encoding paradigms. In the Lamar et al. study [65] the mini-mental state exam (MMSE) did not show any group differences, indicating that the groups do not differ on brain organic state, and thus that result in that study reflects a connection between brain area size and depression score. Cause-effect connections should be explored further, and more detailed regional divisions should be considered.

Learning relies heavily on strategic and executive skills, including mental flexibility and perceptual organization [2] , skills that require alertness and attentional abilities. Older adults, regardless of mental state, are less capable of meta cognitive control over their own learning, since older adults have been demonstrated to be much less effective in restudying items by choice [66] again demonstrating less effective information processing in older adults. Thus, it should be expected that learning ability should be more impaired in older depressed than in younger depressed. Indeed, in their meta study McClintock et al. [2] refer studies which have found significantly higher error rates in late life depression than in non-depressed age matched controls on several learning tests, including the Cambridge Neuropsychological Test Automated Battery (subtest on Visuospatial Paired-Associates Learning), Rey Auditory Verbal Learning Test (RAVLT, verbal list-learning and distractor list), and the Hopkins Verbal Learning Test and the California Verbal Learning Test (CVLT).

Set shifting ability is consistently found to be reduced in affective disorders, including eating disorder [70] [71] , obsessive compulsive disorder [72] and depression [21] . Set shifting and sustained attention to new stimuli are interdependable cognitive processes [32] , and it is challenging to measure “attention” and “set-shifting” separate from each other. Set shifting is often measured with the trails-B test (TMT-B, drawing a line between alternating numbers and letters); [42] and/or the WCST (which rely on several closely linked executive functions, see above), but interpretations are not straight forwards. Impairments on the TMT-B in depressed groups may be best explained by the attentional aspects of that task [30] . Butters et al. [21] found that all cognitive effects in late life depression, including the TMT and WCST were mediated by slowed information processing speed, which may also be seen as an indicator for an underlying attentional deficit in depression. The Stroop test [73] consists of three tasks, either to judge the color of stimuli, read color words, or read color words written in a different color than the color mentioned in the text. This latter test is labeled the Stroop interference test. In the study by Butters et al. [21] , late life depressed patients were impaired on both the TMT-B and WCST when compared with a control group, but not on the Stroop test which primarily is a measure of cognitive flexibility, sustained attention and inhibition, since it measures interference between tasks rather than shifting between them.

Spontaneous flexibility requires an active search for alternatives rather than a reactive adjustment to changing rules. One example is the ability to create multiple words according to specific rules within a limited time, as in the F-A-S test [74] . This test requires the participant to find as many nouns as possible starting with the letter F, A, and S. Performance on this task relies heavily on frontal brain structures, since patients with frontal lobe related deficits often perform poorly on word generation on one or all letters. The ability to actively create words depends on motivation and alertness. Thus, low FAS score correlates with increasing apathy in dementia [75] but does not correlate with apathy in depression [76] even though verbal fluency is often referred to as reduced in depression [2] . The FAS-test and other verbal fluency tests used to measure spontaneous flexibility and verbal fluency all rely on intact processing speed and attentional ability, thus factors related to attention and processing speed may be the principle components for low results of the FAS-test in depression [2] [25] . In group studies, it is important to control for these other factors which may lead to reduced verbal fluency or spontaneous flexibility on the individual level.

Source monitoring imply recollecting how we obtained one piece of information, and may involve different processes depending on whether the source is known to be self-generated or externally generated, or whether it is unknown if the source is self-generated or externally generated [77] . Source monitoring of either self-generated or external sources has been found to be poorer in older than for younger adults [24] . Older adults perform particularly poorer when discriminating between externally generated sources as opposed to self-generated sources [78] . Performance in elderly correlated on a measure of inhibition (Stroop-C interference task)/(Strop A + B)), pointing to poorer inhibition control in elderly as one reason for poorer source monitoring. Another possible explanation for these results is that the more complex interference task (C) require more than the available information processing power than the simple Stroop tasks (A and B), and that the results reflect poorer information processing ability in the older subjects. Source monitoring deficits have also been reported in depression. Ulrich and Meiser [79] found that depressed patients when assessed with the Beck Depression Inventory [80] showed better memory for irrelevant stimuli relative to stimuli required for task performance in a source monitoring experiment. The subjects had to decide if a word was old or new, and if old, if the word had been presented on the left or the right side of the screen, and if the word had been presented inside a green frame or a red frame. Even though the depressed patients performed comparable to the control group on the source monitoring task, the depressed group paid significantly more attention to stimuli that were irrelevant to the task completion, suggesting that defocusing of attentional resources in depression may be important for cognitive effects of depression.

Awareness of making errors is important for performance enhancement and learning to perform a task. When humans commit errors, even when not explicitly aware of it, a massive negative peek can be measured from central and frontal electroence phalography electrodes in experiments tracing event related potentials. The negative peak has been labeled error-related negativity (ERN). Functional Magnetic Resonance Imaging (fMRI) evidence points to the origin of this peak within the anterior cingulate cortex (ACC) in particular [81] [82] and also in the dorsolateral prefrontal cortex (DLPFC) to some degree [13] [83] . The response may be a subconscious response to simple detection of errors, or perhaps to a more complex conflict monitoring system [84] .

Performance monitoring is essential in order to solve certain complex tasks, such as the Stroop interference test. Patients with major depressive disorder (MDD) perform poorly on this type of task. Holms & Pizzagalli [85] reported reduced ERN-peaks in N2 and N450 amplitudes in patients with major depression when making errors, thought to originate ACC and the left DLPFC. The MDD patients that performed poorest on the Stroop interference test exhibited the lowest ERN-peak in dACC following stimulus presentation, indicating that poor conflict monitoring may correspond to poor performance on the Stroop interference test. The cause of the low dACC activation may also be related to conflicting processes related to the depression, which perhaps rely on the dACC or interconnected brain areas. Recent studies have shown that ERN-peak correlate with a wide range of affective and emotional factors [86] [87] .

Georgiadi et al. [88] used a go/no-go task to compare remitted major depressive patients and acutely depressed patients with controls. They reported significantly higher ERN peaks in the remitted patients, and nonsignificant ERN-peak reduction in acutely depressed. They also observed a trend of decreased ERN-peaks with age in all study groups, suggesting that ERN-peak normally is reduced with age, corresponding to reduction in acutely depressed patients as compared to remitted patients. These ERN results suggest that regions within the ACC corresponding to an error monitoring system are abnormally active in remitted depressed patients, while several factors, including aging and acute depression reduces activation in the error monitoring system. The higher ERP-peaks in remitted depressed patients may indicate that depressive periods can cause some permanent effects to the ACC [1] , but it is also possible that these effects were pre-existing in the remitted depressed patients. In that case highly activated ERN-peaks may be an indicator for risk of developing depression.

The ability to control actions and stop inappropriate or erroneous responses is important for a wide range of situations and crucial for efficient task performance and mental processing [13] [89] . Inhibition is not one process, but several conscious and unconscious processes that suppress the irrelevant information and enables focusing and concentration on an important task. Inhibition performance is commonly tapped by the Stroop interference task, but also with several task-switching paradigms, semantic generation tasks with rules which require response inhibition, random number generation which require response monitoring and inhibition of repetition of numbers, the stop signal task (SST) where the participant responds to rapidly presented stimuli but withhold the response when a simultaneous tone is presented, or tasks where the participant must inhibit the logical response. For example Tap 2 from Luria and Homskaya [90] where the participant were asked to tap two times when the experimenter shows them one finger and to tap one time when the experimenter shows them two fingers.

Reduced performance on inhibition control tasks is strongly associated with depression [18] [58] [91] -[94] , and inhibition and inhibition control may be key factors to our understanding of depression. Age is also a contributing factor to reduced inhibition control, and age has been found to predict 30.8% of the variance in inhibition when level of education was controlled for [12] . Thus, it may be assumed that inhibition control is of particular importance in late life depression. De Lissnyder et al. [94] found that depressed patients with rumination symptoms were significantly more impaired in executive control tasks than depressed patients without rumination, suggesting that symptomatology and depression severity should be controlled for in order to assess the role of inhibition control in depression.

Animal research has shown that certain neurotransmitters i.e., acetylcholine may trigger the ACC to enhance inhibition of the lateral prefrontal cortex [95] , and human research indicate that increasing cholinergic activity also may shift human behavior towards inhibition and poorer cognitive control in manic depression and schizophrenia [48] . Integrity of the rostal ACC has been found to be important for cognitive control, including response inhibition when conflicts are being detected and during subsequent evaluation of errors [13] [18] . The functioning of the rostal ACC has been found to be altered due to depression, as shown by deviating ERP responses in depressed older adults following an emotional go―no/go paradigm [93] . Also, smaller gray matter volumes of the ACC correlated with poor effects of antidepressant treatment [96] , indicating that preexisting conditions affecting the anterior cingulate may be one reason why many elderly depressed patients do not respond to common antidepressant treatments [18] . Reduced N2 components within ACC correlate with poorer performance on the inhibition control task, suggesting that depression causes less efficient processing within the ACC, or perhaps that brain plasticity has shifted inhibition control processing to less specialized areas [13] [18] . However, it is not clear whether the mechanisms behind the altered functioning of the ACC may be a direct consequence of depression, or if a deficit within the ACC causes depression.

Katz et al. [18] found that deficit in inhibition control vary as a function of task load. Their data suggested that ERP responses were normal in depressed subjects, but abnormal peaks originating from the ACC occurred in depression patients under high-load conditions only. ERPs which normally show a peak of the N2 component during successful inhibition control is reduced in depressed elderly, accompanied by enhanced N2 components in the middle frontal gyrus [18] . Thus, underlying factors, such as limited capacity in processing due to poorer information processing speed and subsequent attentional factors must be important also for inhibition and inhibition control, since inhibition control is poorer under high task load.

Deficits involving planning may be caused by deficits in a number of underlying functions, such as the ability to attend to a task or insufficient speed of processing. It is clear from previous research that high level executive functioning such as planning particularly relies on the integrity of prefrontal brain regions that also are important for attention, such as the dorsolateral frontal lobe [6] .

Cognitive planning may be assessed with the “Zoo Map Test” which consists of two subtests, one low demanding test where the participants only follow a predefined strategy in order to proceed from one site to the next site in a zoo-map, and one high demanding test where the participants must plan their route according to complex rules [97] . Allain et al. [98] concluded that planning ability is affected in aging when comparing a young and an older age group using the “Zoo Map Test”. They found that the older group exhibited more problems than the younger group when they needed to plan a route than when they just followed a predefined pattern, and the authors suggested that elderly participants may have problems in developing “logical strategies”. Since the “Zoo Map Test” also correlate with processing time, it is possible that group differences in processing time contributes to this finding. Oosterman et al. [99] compared this test against a number of other neuropsychological tests grouped in the domains of “planning”, “memory”, “inhibition” and “processing time” and found the strongest correlation for the “planning” domain. However, the authors note that processing speed also correlate with performance and that processing speed may be the strongest predictor of the “Zoo Map Test” when comparing different patient groups or less homogeneous groups.

Elderkin-Thompson et al. [56] used the “Rey Osterrieth Complex Figure test (ROCF)” when comparing a group of late life depressed with a control group, in order to assess whether poor performance on the in late life depression depends on reduced planning ability or reduced visuospatial ability. They found that poorer scores for a depressed group was quantified by planning simple copying of the ROCF figure and concluded that planning may be the underlying source of declining nonverbal memory in late life depression. Their results also shows that deviation on ROCF-copy scores in late life depression is disassociated from visuospatial ability, but it is possible that the decline in ROCF-copy performance reflect underlying poorer short term memory span due to poorer processing speed and/or attentional ability rather than planning ability.

Aging is an important co-factor which leads to more severe cognitive effects due to depression, thus age related decline in motor functioning, accuracy and speed of processing must be controlled for in group studies [19] [20] . Educational level is another factor that influences cognitive function, and thus most neuropsycological tests controls for these factors. In group studies involving elderly patients, the motor speed components should be controlled for with additional similar tests which do not require the same cognitive functions that are being tested. Also, as reviewed here, several other factors should ideally be controlled for in future studies of the cognitive effect on depression, which may lead to the requirement of larger sample groups than planned for initially. A principal components analysis (e.g. [22] ) is one way of controlling for co-factors.

Depression posing as dementia is common in elderly individuals, presenting with symptoms of a recent cognitive decline. In contrast to Alzheimer’s disease where short-term memory loss occurs as a cardinal first sign, all memory is affected equally in the depressed individual. Comorbid depression and degenerative conditions such as mild cognitive impairment and dementia may cause worsening of the symptoms of the degenerative condition or depression, but may be caused by different mechanisms, since faster than expected reaction times have been shown to cause poor performance on accuracy during information processing in dementia [44] . If depression and dementia are coexisting states, the time of onset of depression may help to pinpoint the etiology of the disorder. The Folstein Mini-Mental State Examination (MMSE) or the Montreal Cognitive Assessment (MoCA) are useful tools used for screening of cognitive deficits. The examiner will also look for signs concerning motivation and sense of reality. Patients with dementia tend to try to answer questions, but often fail, while depressed patients may show passivity and give “don’t know” answers. When interviewing the patients, demented patients may be unaware, or deny having memory problems, while depressed elderly may complain about having memory problems.

Depression state and severity of depression should be controlled for in all group studies, since some cognitive effects may only be measurable in particular depression samples. Some effects of depression, e.g. inhibition control may be correlated to depression severity. In order to make a diagnosis of major depression, DSM-IV and/or ICD-10 definitions for the diagnosis must be met, including depressed mood or loss of interest or pleasure. Severity of depression may be measured on the Montgomery Aasberg Depression Rating Scale (MADRS, [101] ), Beck depression inventory [80] , the Hamilton rating scale for depression [54] , or other Depression Rating Scales. There are several Depression Scales that have been evaluated for several age groups, e.g. the MADRS that has been shown to be valid screening instrument also for depression in elderly patients [102] .

Antidepressant medication may influence cognition in several ways, since some medication may lead to additional slowing due to drowsiness, while eliminating symptoms of depression. Most antidepressant medication will affect alertness and other attentional properties. Thus, care should be taken before comparing medicated and non-medicated groups, and in particular mixed groups with different set of medication.

Hormonal differences between the sexes make it about twice as likely for women than men to develop depression, and the effects of depression may differ in women and in men. Thus, study groups should be balanced between sexes. Some other genic factors may be more difficult to control for in group studies, but it is possible that some genetic predispositions to depression may play a role in the development of cognitive impairment secondary to depression, but the evidence for this is weak [2] . Particular interest has been invested in the ApoE4 allele gene, which is known as a risk factor for Alzheimer’s disease. Butters et al. [21] did not find any cognitive effects, particularly no effects on memory function in depressed elderly with ApoE4, hence there is no known link between the presence of depression and this gene.

Comorbidity with another affective disease with overlapping symptoms may lead to more severe cognitive effects on depression. Comorbid depression and anxiety is particularly common among elderly. In a Norwegian population based study including more than 1300 elderly, 9% had depression, 13% had anxiety, and 5% had both [103] . This overlap should be controlled for, since cognitive effects of anxiety and depression both affect executive speed of processing, language, executive functions and memory functions [2] [104] . Also, both depression and anxiety disorders affect attention, although the cognitive effects are different. In depression, general alertness is biased due to a depressive state, while in anxiety disorders attention is biased towards situations which feels threatening. Little is known about the neural underpinnings that separate anxiety from depression, thus future studies are needed to address this relationship.

Comorbidity of depression and certain somatic conditions, such as vascular disease or cardiovascular risk factors (hypertension, diabetes, coronary artery problems) has a similar effect on cognition [105] . The common denominator is that comorbidity of any of these conditions leads to more severe degradation of the cognitive functions when compared with another depressed group without these comorbid conditions. It is possible that some of these conditions, just like aging, lead to slowing of processing speed, and might lead to an indirect slowing of processing speed due to inattention. Overlapping cognitive effects can also be caused by direct affection of brain areas involved in cognition.

Since the cognitive effects of depression overlaps with most, if not all of these possible co-factors, it is challenging to distinguish between the origins of the cognitive effects when a patient has a comorbid condition overlapping depression. However, the evidence points to a worsening of perhaps all the cognitive functions affected in depression if any of these co-factors are present.

Information processing speed is a key factor when addressing the effect of aging and depression, and the combination of aging and depression seems to amplify cognitive symptoms. It is a particular function related to attentional ability that is directly affected, while most cognitive functions that require alertness and attention are secondarily affected. Inhibition is one executive function which is particularly associated with depression, where there are independent findings which suggest that this function may be at the core of understanding the nature of depression [13] [18] .

Neuropsychological tests are useful in assessing depression in aging, but may not predict depression without access to premorbid information and additional information about the subject. Tests that are particularly sensitive to cognitive delay in depressed elderly patents are identified in the present review. In order to address specific questions, particularly designed laboratory test should be designed, or the combination of cognitive test should be carefully considered to ensure that they address the particular research question, and that other effects may be accounted for. Thus, in group studies, it is important to include tests that are targeting a specific cognitive function or sub-function. Several neuropsychological tests are therefore not appropriate for use in group studies because poor scores on a group level may be due to several different individual subject factors, and should be controlled at the individual level in order to enable prediction to which sub-functions are impaired. A systematic approach to assessing cognitive decline due to depression is needed.

Even though attention and inhibition are closely linked cognitive abilities, their relationship should be investigated further. It is not clear whether depression affects inhibition ability directly or if poor inhibition in depression is due to reduced information processing speed and attentional ability because of conflicting negative thought which is a consequence of the depression. This relationship may be studied with fMRI or other imaging techniques in combination with neuropsychological tests or tailored laboratory tests which measure inhibition ability, attention to stimuli and processing speed separately, or include controls for each of these factors. The study groups should be carefully selected so that they are as homogeneous as possible, with respect to depression state and severity, medication, and diagnose, or these factors should be controlled for. Control group needs to be matched for education and age, and perhaps also other co-factors as discussed above.