The aim of the study was to elucidate which parts of the psychiatric interview in patients with depression are most informative regarding night sleep and alertness in the daytime. To validate the patients' subjective estimates, their correlations to the polygraphic indicators of sleep and alertness were evaluated. Trying to interpret the results, the main impression was that many of the subjective complaints could not be verified by the results of objective measurement. If these findings are generally valid, the patients' statements on sleep quality and alertness in the daytime, as routinely examined by a psychiatric interview, are a poor instrument to diagnose this type of symptom. In particular, the statements on the type of insomnia, the estimated time of falling asleep, frequent awakenings and occurrence of disturbing dreams seem to be unreliable. These results are compatible with previosly published reports on differences between subjective and objective statements. However, the discrepancies seem to be more pronounced in the present study. A possible explanation can be the fact, that Argyrpoulos et al.  and Tsuchiyama  studied patients with major depression while the present study was based on the observations in patients with minor depression. Another confounding factor can be the wide age range as can be seen in the present study. As shown by Lee et al. , the subjective assessment of the sleep quality in depression is age-dependent. To estimate the possible influence of various ages in the present study, the statistical processing was enlarged by calculation of partial correlation coefficients, with the age as background variable. However, the results were not substantially changes as compared with the original correlation coefficients. Similarly, the factor of medication was examined by calculating partial correlations with negative results, too.
The results were also disappointing when the patients were asked about alertness disturbances in the daytime. An unexpected finding was the lack of any significant correlation to the scores obtained by means of Epworth's scale, which is generally considered to be an efficient tool in clinical work and research. Of course, the relevance of the individual items varied and some important correlations could be verified. Positive results were also seen in some of the patients' estimates of alertness disturbances in the daytime where simple questions on the feeling of refreshment after sleep, low working capacity, unpleasantness and anxiety gave rather good information.
Besides the age and medication, it was also of interest to study other interfering factors which possibly could influence the patient's alertness and sleep. Among them, alcohol intake seemed to be of more importance than nicotine or coffee intake. On the other hand, the statistical evaluation confirmed that anxiety dominate among possible sources of sleep and alertness disturbances in psychiatric patients .
The generally low correlation between the subjective scores and objective in4dicators necessarily evokes a question on the adequacy of the methods applied. As regards the subjective scores, these were obtained during a structured interview, using standardised questions. The symptoms were registered in a semiquantitative form, as a score on an arbitrary five-point scale, or by means of commonly accepted scales (Beck, Epworth).
Doubts can probably be directed against the adeqacy of polygraphic recording. The representativeness of an examination done on one or two occasions under artificial laboratory conditions can be questioned. However, the technical circumstances do not allow us to adapt the patient by repeating the examination during several nights and the discomfort due to the recording equipment cannot be avoided whatever technique is used. A possible solution would be to make the recording at home, using a portable device. This is certainly available, if only a simple excamination, for example by registering movements during sleep, is desired. However, we did not find a way to get a complete polygram of good quality under such conditions. After all, there is an argument in favour of the technique as it has been used in the study: the patients' estimates of sleep disturbances caused by the examination procedure were not significantly correlated to any of the objective indicators. Thus, no association between sleep quality and patient's discomfort could be proved. The objection that a measurement during a single night does not represent the sleep quality during a longer period is weakened by the fact that the patients' "sleep calender" showed rather stable conditions during the last 2 weeks, including the night of polygraphy.
Whole-night polygraphy is a method which is widely used in sleep laboratories for diagnostic purposes. The evaluation of the recordings is based on definitions of sleep stages well established since fifties . The technique, scoring and terminology, as well as a set of sleep indicators, are standardised . Besides the proportion of the individual sleep stages, several other measures such as sleep latency and sleep efficacy are used (see Methods). The sleep indicators were originally developed for diagnostic purposes, in patients with narcolepsy and apnoic syndrome. It was considered of interest to judge the usefulness of the individual indicators in psychiatric disorders. Some information on that can be obtained from the number of significant correlations in Tables II and IV. As it will be seen, the sleep efficacy, or its modification (SPT/TIB), seem to be more informative than the other polysomnographic indicators.
In contrast to night sleep, the methods of examining the alertness level in the daytime are more disputable. In fact, neither the clinical terminology or every-day language is quite clear in describing various symptoms. For example, as shown in a previous study on patients with mild dementia , the expression "fatigue" correlated with the degree of the cognitive defect, without any significant correlation with the sleep and alertness indicators. In the present study, the difficulties are well illustrated by the disappointing results achieved with Epworth's scale. This scale is well established and widely applied in examining the symptoms of sleepiness in the daytime  but the scores did not show any relevant correlation to any of the objective indicators. The relevance of the objective indicators could, however, be supported by their significant correlations to other clinical items (Tables III and IV). Moreover, interpretation of the visual indicators is not difficult because they are based, similarly to the indicators of sleep quality, on conventional sleep stages. The only difference is that the assessment has been refined by using several sub-stages in order to register the transitions between full wakefulness (sleep stage 0) and decreased alertness (sleep stage 1). These transitions are characterised by rather typical EEG patterns  and can be defined as typical sub-stages. The method has previously been used in a study on healthy individuals and the significant correlation between the EEG pattern and a subjective score for sleepiness in the daytime could be confirmed . In the previous studies, it was also found that the method was more sensitive in this application than the Multiple Sleep Latency Test, a method of choice for examining patients with narcolepsy . Another procedure, the Maintance of Wakefulness Test , is probably more sensitive but it was not tested. The reason was that it requires a degree of cooperation that could not be achieved in psychiatric patients with pronounced anxiety symptoms.
To get another, more accurate indicator than the visual scores, the subtle transitions between sleep stages 0 and 1 were also assessed by means of spectrum analysis. Because the shift from full alertness to sleep 1 stage is mainly characterised by successive disappearence of the dominant alpha activity , its amplitude in the individual epochs could be used as an indicator. This type of measurement, based on the calculation of EEG spectra, belongs to relatively well established methods of EEG analysis . Thus, the application of spectrum analysis to estimate the transitions from wakefulness is acceptable and not controversial . More interesting can be the additional data processing, trying to characterise the whole 10 min recording from a set of spectral values as obtained in the individual 10 s epochs. Among various measures, the trend was found to be most contributive. The unidirectional shift in the course of the examination session was presented by means of regression to the time from the start of the recording. A negative regression coefficient means that the alpha amplitudes showed a tendency toward successive decrease. According to the definitions of the sleep stages, this change can be interpreted as a shift from full alertness (sleep stage 0) toward decreased alertness (sleep stage 1). In the present study, the trend was used for the first time as a measure of alertness shifts and the results, in terms of significant correlations to the clinical items, are encouraging. Therefore, the trend was also calculated for the former indicator, based on visual analysis of the curves, and the results are favourable, too (Tables III and IV).
The main objection against the results of the study concerns the great number of correlation coefficients, each of them representing a statistical test. Considering the limited material available, there is a risk of false positive results. Unfortunately, almost all methods of EEG analysis result in such a large numer of variables that the usual statistical methods to compensate for the "alpha overflow" (e.g. Bonferoni method) cannot be used. The present study certainly belongs to the "exploratory" investigations, requiring a continued "confirmatory" investigation . In any case, the findings can be used to establish new hypotheses.
In conclusion, the study confirms that self-estimation of sleep quality is unreliable in depressed patients. Similarly, there is a discrepancy between the subjective and objective assessment of the alertness changes in the daytime. Of course, the results of the study should not be over-interpreted and the diagnosis should always be based on the clinical examination. However, some kind of "correcting factor" can be considered in this type of patients, particularly in those suffering of anxiety symptoms. The study also demonstrates possible difficulties caused by vague definitions of sleep and alertness disturbances. In uncertain cases, a qualified polygraphic examination is well motivated, anyway.