Conference Paper

Not Just a Matter of Accuracy: a fNIRS Pilot Study into Discrepancy between Sleep Data and Subjective Sleep Experience in Quantified-Self Sleep Tracking

Authors:
  • Kyoto University of Advanced Science
To read the full-text of this research, you can request a copy directly from the author.

Abstract

Quantified-self sleep tracking devices such as Fitbit are becoming increasingly popular in recent years. However, users often complain about the discrepancy between the data collected with sleep trackers and their subjective sleep experience, which is often attributed to the accuracy issue of the devices. In this pilot study, we aim to provide an explanation to such discrepancy from a neuroscience perspective. We investigate the associations of subjective sleep rating and Fitbit measured sleep data to cortical hemodynamics in the prefrontal cortex (PFC) during the first sleep cycle. Correlation analysis results show that subjective sleep rating mainly correlates to the median of the concentration changes in oxyhemoglobin (ΔO2Hb) and deoxyhemoglobin (ΔHHb) in a set of channels, with positive correlation coefficients. In contrast, the sleep score computed by Fitbit mainly correlates to the mean of the ΔO2Hb and ΔHHb in a different set of channels, with negative correlation coefficients. This suggests that better perceived sleep quality may be positively associated to increased hemodynamics during the first sleep cycle, and the opposite is true for objective sleep metrics such as sleep score measured by Fitbit. The result implies that users' subjective perception of sleep and the sleep tracking devices may be capturing different dimensions of sleep. As such, improving device accuracy may help little in addressing the discrepancy between subjective sleep experience and objective data. Based on the finding, we point out directions for future studies on sleep tracking technologies.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the author.

Article
While data is the cornerstone of modern design strategies, design researchers frequently struggle when performing data work. This creates a need to design tools that enable design researchers to actively engage with data. However, this presupposes understanding how design researchers create meaning from data representations, as the way of visualizing the data, along with other factors, can significantly impact the extracted insights, increasing uncertainty about the quality of the outcome. As a response to this problem, we explore how design researchers make sense of data in a case study: making sense of paired subjective and objective sleep and stress data visualizations. By synthesizing our findings from two user studies, we construct a sensemaking model which highlights how uncertainty related to data qualities, visualization parameters, and the viewer’s background, affects the insight-generation process. Our findings have implications for the future development of tools and techniques for visual data sensemaking for designers.
Article
Full-text available
Consumer wearable activity trackers such as Fitbit are widely used in ubiquitous and longitudinal sleep monitoring in free-living environments. However, these devices are known to be inaccurate for measuring sleep stages. In this study, we develop and validate a novel approach that leverages the processed data readily available from consumer activity trackers (i.e., steps, heart rate, and sleep metrics) to predict sleep stages. The proposed approach adopts a selective correction strategy and consists of two levels of classifiers. The level-I classifier judges whether a Fitbit labelled sleep epoch is misclassified, and the level-II classifier reclassifies misclassfied epochs into one of the four sleep stages (i.e., light sleep, deep sleep, REM sleep and wakefulness). Best epoch-wise performance was achieved when support vector machine and gradient boosting decision tree (XGBoost) with up sampling were used respectively at the level-I and level-II classification. The model achieved an overall per-epoch accuracy of 0.731 ± 0.119, Cohen's Kappa of 0.433 ± 0.212, and multi-class Matthew's correlation coefficient (MMCC) of 0.451 ± 0.214. Regarding the total duration of individual sleep stage, the mean normalized absolute bias (MAB) of this model was 0.469, which is a 23.9% reduction against the proprietary Fitbit algorithm. The model that combines support vector machine and XGBoost with down sampling achieved sub-optimal per-epoch accuracy of 0.704 ± 0.097, Cohen's Kappa of 0.427 ± 0.178, and MMCC of 0.439 ± 0.180. The sub-optimal model obtained a MAB of 0.179, a significantly reduction of 71.0% compared to the proprietary Fitbit algorithm. We highlight the challenges in machine learning based sleep stage prediction with consumer wearables, and suggest directions for future research.
Article
Full-text available
Consumer sleep-tracking devices provide an unobtrusive and affordable way to learn about personal sleep habits. Recent research focused primarily on the information provided by such devices, i.e., whether the information is accurate and meaningful to people. However, little is known about how people judge the credibility of such information, and how the functionality and the design may influence such judgements. Hence, the aim of this research was to examine how consumers assess the credibility of sleep-tracking devices. We conducted a qualitative study with 22 participants who tracked their sleep for 3 nights with three different devices: Fitbit Charge 2, Neuroon EEG, and SleepScope, a medical sleep monitor. Based on semi-structured interviews, we found that people assess the credibility of sleep-tracking devices based not only on the credibility of sleep data per se, but also on device functionality, interface design and physical appearance. People found it difficult to judge credibility, because of the complexities of sleep stages and micro-arousals (sleep fallacy) and the black boxed nature of devices (black box fallacy), and also because of the misalignment between objective sleep measures and subjective sleep quality. We discuss the significance of design and functionality on the credibility of personal health technologies and highlight design challenges and opportunities to enhance their credibility.
Article
Full-text available
Background: It has become possible for the new generation of consumer wristbands to classify sleep stages based on multisensory data. Several studies have validated the accuracy of one of the latest models, that is, Fitbit Charge 2, in measuring polysomnographic parameters, including total sleep time, wake time, sleep efficiency (SE), and the ratio of each sleep stage. Nevertheless, its accuracy in measuring sleep stage transitions remains unknown. Objective: This study aimed to examine the accuracy of Fitbit Charge 2 in measuring transition probabilities among wake, light sleep, deep sleep, and rapid eye movement (REM) sleep under free-living conditions. The secondary goal was to investigate the effect of user-specific factors, including demographic information and sleep pattern on measurement accuracy. Methods: A Fitbit Charge 2 and a medical device were used concurrently to measure a whole night's sleep in participants' homes. Sleep stage transition probabilities were derived from sleep hypnograms. Measurement errors were obtained by comparing the data obtained by Fitbit with those obtained by the medical device. Paired 2-tailed t test and Bland-Altman plots were used to examine the agreement of Fitbit to the medical device. Wilcoxon signed-rank test was performed to investigate the effect of user-specific factors. Results: Sleep data were collected from 23 participants. Sleep stage transition probabilities measured by Fitbit Charge 2 significantly deviated from those measured by the medical device, except for the transition probability from deep sleep to wake, from light sleep to REM sleep, and the probability of staying in REM sleep. Bland-Altman plots demonstrated that systematic bias ranged from 0% to 60%. Fitbit had the tendency of overestimating the probability of staying in a sleep stage while underestimating the probability of transiting to another stage. SE>90% (P=.047) was associated with significant increase in measurement error. Pittsburgh sleep quality index (PSQI)<5 and wake after sleep onset (WASO)<30 min could be associated to significantly decreased or increased errors, depending on the outcome sleep metrics. Conclusions: Our analysis shows that Fitbit Charge 2 underestimated sleep stage transition dynamics compared with the medical device. Device accuracy may be significantly affected by perceived sleep quality (PSQI), WASO, and SE.
Article
Full-text available
Background Few studies assessing the correlation between patient-reported outcomes and patient-generated health data from wearable devices exist. Objective The aim of this study was to determine the direction and magnitude of associations between patient-generated health data (from the Fitbit Charge HR) and patient-reported outcomes for sleep patterns and physical activity in patients with type 2 diabetes mellitus (T2DM). Methods This was a pilot study conducted with adults diagnosed with T2DM (n=86). All participants wore a Fitbit Charge HR for 14 consecutive days and completed internet-based surveys at 3 time points: day 1, day 7, and day 14. Patient-generated health data included minutes asleep and number of steps taken. Questionnaires assessed the number of days of exercise and nights of sleep problems per week. Means and SDs were calculated for all data, and Pearson correlations were used to examine associations between patient-reported outcomes and patient-generated health data. All respondents provided informed consent before participating. ResultsThe participants were predominantly middle-aged (mean 54.3, SD 13.3 years), white (80/86, 93%), and female (50/86, 58%). Use of oral T2DM medication correlated with the number of mean steps taken (r=.35, P=.001), whereas being unaware of the glycated hemoglobin level correlated with the number of minutes asleep (r=−.24, P=.04). On the basis of the Fitbit data, participants walked an average of 4955 steps and slept 6.7 hours per day. They self-reported an average of 2.0 days of exercise and 2.3 nights of sleep problems per week. The association between the number of days exercised and steps walked was strong (r=.60, P
Article
Full-text available
Consumersleep trackingtechnologies offer anunobtrusive and cost-efficient waytomonitor sleepinfree-livingconditions.Technologicaladvancesinhardwareand software have significantly improved the functionality of the new gadgets that recently appeared in the market. However, whether the latest gadgets can provide valid measurements on overall sleep parameters and sleep structure such as deep and REM sleep has not been examined. In this study, we aimed to investigate the validity of the latest consumer sleep tracking devices including an activity wristband Fitbit Charge 2 and a wearable EEG-based eye mask Neuroon in comparison to a medical sleep monitor. First, we confirmed that Fitbit Charge 2 can automatically detect the onset and offset of sleep with reasonable accuracy. Second, analysis found that both consumer devices produced comparable results in measuring total sleep duration and sleep efficiency compared to the medical device. In addition, Fitbit accurately measured the number of awakenings, while Neuroon with good signal quality had satisfactory performance on total awake time and sleep onset latency. However, measuring sleep structure including light, deep, and REM sleep remains to be challenging for both consumer devices. Third,greater discrepancies were observed between Neuroon and the medical device in nights with more disrupted sleep and when the signal quality was poor, but no trend was observed in Fitbit Charge 2. This study suggests that current consumer sleep tracking technologies may be immature for diagnosing sleep disorders, but they are reasonably satisfactory for general purpose and non-clinical use.
Article
Full-text available
Background Single-patient, multiple cross-over designs (N-of-1 or single-case randomized clinical trials) with systematic data collection on treatment effects may be useful for increasing the precision of treatments in health psychology. Purposes To assess the quality of the methods and statistics, describe interventions and outcomes, and explore the heterogeneity of treatment effect of health psychology N-of-1 trials. Methods We conducted a systematic review of N-of-1 trials from electronic database inception through June 1, 2015. Potentially relevant articles were identified by searching the biomedical electronic databases Ovid, MEDLINE, EMBASE, all six databases in the Cochrane Library, CINAHL, and PsycINFO, and conference proceedings, dissertations, ongoing studies, Open Grey, and the New York Academy’s Grey Literature Report. Studies were included if they had health behavior or psychological outcomes and the order of interventions was randomized. We abstracted study characteristics and analytic methods and used the Consolidated Standards of Reporting Trials extension for reporting N-of-1 trials as a quality checklist. Results Fifty-four N-of-1 trial publications composed of 1,193 participants were included. Less than half of these (36%) reported adequate information to calculate the heterogeneity of treatment effect. Nearly all (90%) provided some quantitative information to determine the superior treatment; 79% used an a priori statistical cutoff, 12% used a graph, and 10% used a combination. Conclusions N-of-1 randomized trials could be the next major advance in health psychology for precision therapeutics. However, they must be conducted with more methodologic and statistical rigor and must be transparently and fully reported.
Article
Full-text available
Background: Smart wearables such as the Fitbit wristband provide the opportunity to monitor patients more comprehensively, to track patients in a fashion that more closely follows the contours of their lives, and to derive a more complete dataset that enables precision medicine. However, the utility and efficacy of using wearable devices to monitor adolescent patients' asthma outcomes have not been established. Objective: The objective of this study was to explore the association between self‑reported sleep data, Fitbit sleep and physical activity data, and pediatric asthma impact (PAI). Methods: We conducted an 8‑week pilot study with 22 adolescent asthma patients to collect: (1) weekly or biweekly patient‑reported data using the Patient-Reported Outcomes Measurement Information System (PROMIS) measures of PAI, sleep disturbance (SD), and sleep‑related impairment (SRI) and (2) real-time Fitbit (ie, Fitbit Charge HR) data on physical activity (F-AM) and sleep quality (F‑SQ). To explore the relationship among the self-reported and Fitbit measures, we computed weekly Pearson correlations among these variables of interest. Results: We have shown that the Fitbit-derived sleep quality F-SQ measure has a moderate correlation with the PROMIS SD score (average r=-.31, P=.01) and a weak but significant correlation with the PROMIS PAI score (average r=-.18, P=.02). The Fitbit physical activity measure has a negligible correlation with PAI (average r=.04, P=.62). Conclusions: Our findings support the potential of using wrist-worn devices to continuously monitor two important factors-physical activity and sleep-associated with patients' asthma outcomes and to develop a personalized asthma management platform.
Article
Full-text available
Background: Reports of subjective sleep quality are frequently collected in research and clinical practice. It is unclear, however, how well polysomnographic measures of sleep correlate with subjective reports of prior-night sleep quality in elderly men and women. Furthermore, the relative importance of various polysomnographic, demographic and clinical characteristics in predicting subjective sleep quality is not known. We sought to determine the correlates of subjective sleep quality in older adults using more recently developed machine learning algorithms that are suitable for selecting and ranking important variables. Methods: Community-dwelling older men (n=1024) and women (n=459), a subset of those participating in the Osteoporotic Fractures in Men study and the Study of Osteoporotic Fractures study, respectively, completed a single night of at-home polysomnographic recording of sleep followed by a set of morning questions concerning the prior night's sleep quality. Questionnaires concerning demographics and psychological characteristics were also collected prior to the overnight recording and entered into multivariable models. Two machine learning algorithms, lasso penalized regression and random forests, determined variable selection and the ordering of variable importance separately for men and women. Results: Thirty-eight sleep, demographic and clinical correlates of sleep quality were considered. Together, these multivariable models explained only 11-17% of the variance in predicting subjective sleep quality. Objective sleep efficiency emerged as the strongest correlate of subjective sleep quality across all models, and across both sexes. Greater total sleep time and sleep stage transitions were also significant objective correlates of subjective sleep quality. The amount of slow wave sleep obtained was not determined to be important. Conclusions: Overall, the commonly obtained measures of polysomnographically-defined sleep contributed little to subjective ratings of prior-night sleep quality. Though they explained relatively little of the variance, sleep efficiency, total sleep time and sleep stage transitions were among the most important objective correlates.
Article
Full-text available
Getting enough quality sleep is a key part of a healthy lifestyle. Many people are tracking their sleep through mobile and wearable technology, together with contextual information that may influence sleep quality, like exercise, diet, and stress. However, there is limited support to help people make sense of this wealth of data, i.e., to explore the relationship between sleep data and contextual data. We strive to bridge this gap between sleep-tracking and sense-making through the design of SleepExplorer, a web-based tool that helps individuals understand sleep quality through multi-dimensional sleep structure and explore correlations between sleep data and contextual information. Based on a two-week field study with 12 participants, this paper offers a rich understanding on how technology can support sense-making on personal sleep data: SleepExplorer organizes a flux of sleep data into sleep structure, guides sleep-tracking activities, highlights connections between sleep and contributing factors, and supports individuals in taking actions. We discuss challenges and opportunities to inform the work of researchers and designers creating data-driven health and well-being applications.
Article
Full-text available
To examine the role of objective sleep duration, a novel marker in phenotyping insomnia, and psychological profiles on sleep misperception in a large, general population sample. Sleep misperception is considered by some investigators a common characteristic of chronic insomnia, whereas others propose it as a separate diagnosis. The frequency and the determinants of sleep misperception in general population samples are unknown. A total of 142 insomniacs and 724 controls selected from a general random sample of 1,741 individuals (aged ≥20 years) underwent a polysomnographic evaluation, completed the Minnesota Multiphasic Personality Inventory-2, and were split into two groups based on their objective sleep duration: "normal sleep duration" (≥6 hours) and "short sleep duration" (<6 hours). The discrepancy between subjective and objective sleep duration was determined by two independent factors. Short sleepers reported more sleep than they objectively had, and insomniacs reported less sleep than controls with similar objective sleep duration. The additive effect of these two factors resulted in underestimation only in insomniacs with normal sleep duration. Insomniacs with normal sleep duration showed a Minnesota Multiphasic Personality Inventory-2 profile of high depression and anxiety and low ego strength, whereas insomniacs with short sleep duration showed a profile of a medical disorder. Underestimation of sleep duration is prevalent among insomniacs with objective normal sleep duration. Anxious-ruminative traits and poor resources for coping with stress seem to mediate the underestimation of sleep duration. These data further support the validity and clinical utility of objective sleep measures in phenotyping insomnia.
Conference Paper
The recent rise of the Quantified Self movement has witnessed a significant increase in the adoption of consumer wearable wristbands for sleep tracking. Nevertheless, data quality of these devices has been a main concern. This study aimed to validate a most popular consumer wristband, i.e. Fitbit Charge 2, against medical devices. We proposed a new validation approach that combines numerical technique with visual aid for epoch-by-epoch comparison on sleep stages. We found that Fitbit Charge 2TM had low accuracy in detecting wake and reasonable accuracy in detecting light, deep, and REM sleep stages. The visual aid of scatter plots showed that Fitbit was more accurate in detecting deep sleep stage in the first half of a night and more accurate in detecting REM sleep stage in the second half of a night. Our results indicate that consumer wearable wristbands are not able to produce high quality data of sleep stages in ecological settings. Future studies should consider the effect of time on device accuracy and may resort to segmented modelling techniques to improve data quality.
Article
Subjective perception of sleep is not necessarily consistent with EEG indications of sleep. The mismatch between subjective reports and objective measures is often referred to as "sleep state misperception" (SSM). Previous studies evince that this mismatch is found in both insomnia patients and in normal sleepers, but the neurophysiological mechanism remains unclear. The aim of the study is to explore the neurophysiological basis of this mechanism, from the perspective of both EEG power and fMRI fluctuations. Thirty-six healthy young adults participated in the study. Simultaneous EEG and fMRI recordings were conducted while the participants were trying to fall asleep in an MRI scanner at approximately 9:00 PM. They were awakened after achieving stable N1 or N2 sleep, or after 90 minutes without falling into stable sleep. Next they were asked to recall their conscious experiences from the moment immediately prior to awakening. Sixty-one instances of scheduled awakenings were collected: twenty-nine of these after having achieved stable stage N2 sleep; twelve, during stage N1 sleep; and, twenty during the waking state. Relative to those awakenings without subjecitve-objective discrepancy, these awakenings with discrepancy were associated with lower theta power, as well as higher alpha, beta, and gamma power. Moroever, we found that participants who exhibited the discrepancy, compared to those who did not, evinced a higher amplitude of low-frequency fluctuation (ALFF) levels in the prefrontal cortex. These results lend support to the conjecture that the subjective-objectve discrepancy is associated with CNS hyperarousal. The discrepancy between subjective sleep perception and objective measures of sleep, has been reported commonly in normal sleepers; it is also a clinical feature of some insomnia patients. The underlying mechanism of this phenomenon, however, has not been well studied. This research is the first simultaneous EEG and fMRI attempt to investigate the neurophysiological mechanisms associated with this discrepancy. Results suggest that subjective-objective difference can be explained, in part, by reference to general hyperarousal of the brain, especially along the fronto-parietal pathway that is related to to executive control.
Article
We evaluated the performance of a consumer multi-sensory wristband (Fitbit Charge 2™), against polysomnography (PSG) in measuring sleep/wake state and sleep stage composition in healthy adults. In-lab PSG and Fitbit Charge 2™ data were obtained from a single overnight recording at the SRI Human Sleep Research Laboratory in 44 adults (19—61 years; 26 women; 25 Caucasian). Participants were screened to be free from mental and medical conditions. Presence of sleep disorders was evaluated with clinical PSG. PSG findings indicated periodic limb movement of sleep (PLMS, > 15/h) in nine participants, who were analyzed separately from the main group (n = 35). PSG and Fitbit Charge 2™ sleep data were compared using paired t-tests, Bland–Altman plots, and epoch-by-epoch (EBE) analysis. In the main group, Fitbit Charge 2™ showed 0.96 sensitivity (accuracy to detect sleep), 0.61 specificity (accuracy to detect wake), 0.81 accuracy in detecting N1+N2 sleep (“light sleep”), 0.49 accuracy in detecting N3 sleep (“deep sleep”), and 0.74 accuracy in detecting rapid-eye-movement (REM) sleep. Fitbit Charge 2™ significantly (p < 0.05) overestimated PSG TST by 9 min, N1+N2 sleep by 34 min, and underestimated PSG SOL by 4 min and N3 sleep by 24 min. PSG and Fitbit Charge 2™ outcomes did not differ for WASO and time spent in REM sleep. No more than two participants fell outside the Bland–Altman agreement limits for all sleep measures. Fitbit Charge 2™ correctly identified 82% of PSG-defined non-REM–REM sleep cycles across the night. Similar outcomes were found for the PLMS group. Fitbit Charge 2™ shows promise in detecting sleep-wake states and sleep stage composition relative to gold standard PSG, particularly in the estimation of REM sleep, but with limitations in N3 detection. Fitbit Charge 2™ accuracy and reliability need to be further investigated in different settings (at-home, multiple nights) and in different populations in which sleep composition is known to vary (adolescents, elderly, patients with sleep disorders).
Article
The PDF version is freely available on the internet on https://www.thieme-connect.de/products/ejournals/abstract/10.3414/ME17-03-0001
Article
Good sleep is essential to good health. Yet for most of its history, sleep medicine has focused on the definition, identification, and treatment of sleep problems. Sleep health is a term that is infrequently used and even less frequently defined. It is time for us to change this. Indeed, pressures in the research, clinical, and regulatory environments require that we do so. The health of populations is increasingly defined by positive attributes such as wellness, performance, and adaptation, and not merely by the absence of disease. Sleep health can be defined in such terms. Empirical data demonstrate several dimensions of sleep that are related to health outcomes, and that can be measured with self-report and objective methods. One suggested definition of sleep health and a description of self-report items for measuring it are provided as examples. The concept of sleep health synergizes with other health care agendas, such as empowering individuals and communities, improving population health, and reducing health care costs. Promoting sleep health also offers the field of sleep medicine new research and clinical opportunities. In this sense, defining sleep health is vital not only to the health of populations and individuals, but also to the health of sleep medicine itself. Buysse DJ. Sleep health: can we define it? Does it matter? SLEEP 2014;37(1):9-17.
Article
Twenty healthy men and women had their sleep recorded objectively, using polysomnography on 3 nonconsecutive nights. Following each night, the subjects assessed their sleep onset latency and number of awakenings, subjectively. Self-ratings were compared with objective measures of sleep onset latency (SOL), calculated as the time from lights-out to the first continuous minute of stage 2 sleep, and the number of awakenings which lasted 1 minute or longer on the polysomnograms. Apart from the first night, the subjects overestimated the time that it took them to fall asleep, despite sleep onset being scored as the latency to stage 2, rather than stage 1 sleep. On all 3 nights, the subjects underestimated the number of awakenings when compared to objective criteria. Although the subjects were consistent in their errors of estimation of their sleep compared to polysomnographic assessments over the three nights, the between-individual variation was large, so that objective and subjective ratings of SOL and awakenings were not correlated. The young men and women in our study, who were free of medication or sleep complaints, perceived their sleep inaccurately when compared to objective polysomnographic recordings.
Article
Despite the prevalence of sleep complaints among psychiatric patients, few questionnaires have been specifically designed to measure sleep quality in clinical populations. The Pittsburgh Sleep Quality Index (PSQI) is a self-rated questionnaire which assesses sleep quality and disturbances over a 1-month time interval. Nineteen individual items generate seven "component" scores: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction. The sum of scores for these seven components yields one global score. Clinical and clinimetric properties of the PSQI were assessed over an 18-month period with "good" sleepers (healthy subjects, n = 52) and "poor" sleepers (depressed patients, n = 54; sleep-disorder patients, n = 62). Acceptable measures of internal homogeneity, consistency (test-retest reliability), and validity were obtained. A global PSQI score greater than 5 yielded a diagnostic sensitivity of 89.6% and specificity of 86.5% (kappa = 0.75, p less than 0.001) in distinguishing good and poor sleepers. The clinimetric and clinical properties of the PSQI suggest its utility both in psychiatric clinical practice and research activities.
Ploderer: Sleep tracking in the real world: a qualitative study into barriers for improving sleep
  • Z Liang
Liang, Z., B. Ploderer: Sleep tracking in the real world: a qualitative study into barriers for improving sleep, in Proceedings of the 28th Australian Conference on Computer-Human Interaction, Launceston, Tasmania, Australia. p. 537-541 (2016).
Estimation of optical pathlength through tissue from direct time of flight measurement
  • D T Delpy
Delpy, D.T., et al.: Estimation of optical pathlength through tissue from direct time of flight measurement. Phys. Med. Biol. 33, 1433-1442 (1988).