J . Sleep Res.

(1992) 1, 35-39

Psychometric evaluation of the Stanford Sleepiness Scale

ALISTAIR W. MACLEAN.2, G . CYNTHIA FEKKEN,
PAUL SASKIN2* and JOHN B. KNOWLES
Department of Psychology, Queens University, Kingston, Ontario; Department of Psychiatry, The Toronto Hospital (Western
Division), Toronto, Ontario, Canada;

Accepted in revised form 10 December 1991; revised 12 Novembcr 1991

SUMMARY T w o assumptions underlying the Stanford Sleepiness Scale (SSS) were evaluated:
that the descriptors defining each level of the scale are equivalent ways of
characterizing a particular level of sleepiness; and that sleepiness, thus measured, is
an unidimensional construct. Twenty-four True/False items were derived from t h e
descriptors at each level of the SSS. This revised scale was administered to 340
undergraduates in a questionnaire which also included: t h e SSS; four visual
analogue scales; items identifying the subjects age, sex, a n d circadian type; and the
time of administration. Analyses of the responses indicated that endorsement of
items o n the revised scale was not consistent with the SSS level endorsed, indicating
that the descriptors at each scale level a r e not equivalent. A principal components
analysis revealed two components, tentatively identified as activation and sleepiness,
accounting, respectively, for 24.2 and 20.6% of the variance. It was concluded that
sleepiness is not a n unidimensional construct. Further studies a r e necessary t o
elucidate the nature of its components.
KEYWORDS psychometric evaluation, Stanford Sleepiness Scale, sleepiness

with objective measures in some circumstances has been

INTRODUCTION
The Stanford Sleepiness Scale (SSS-Hoddes et al. 1972,
1973) is probably the most widely used instrument for the
assessment of subjective sleepiness. The only report on the
reliability of the scale indicates that the correlation between
alternate forms is 0.88 (Hoddes et al. 1972, 1973). Attempts
to validate the SSS, with varying degrees of success, have
been made using: studies of sleep deprivation (Hoddes et al.
1973; Glenville and Broughton 1979; Herscovitch and
Broughton 1981a); correlation with performance tests
(Hoddes et al. 1973; Glenville and Broughton 1979;
Herscovitch and Broughton 1981b) and with polysom-
nographically defined sleep onset latency (Dement 1976);
long-term sleep restriction (Friedman et al. 1977); and its
application to clinical samples (Dement et al. 1978; Valley
and Broughton 1981). While it has been generally deemed
to have acceptable levels of validity, its poor correlation
*Present address: Sunnyview Hospital and Rehabilitation Centre,
1270 Belmont Avenue, Schenectady, NY 12308, USA.
Correspondence: Dr A. W . MacLean, Department of Psychology,
Queens University, Kingston, Ontario, Canada K7L 3N6.
commented upon (Dement et al. 1987; Herscovitch and
Broughton 1981b).
Our interest in the SSS was first triggered by the
observation in a study of mood following a phase shift of the
sleep wake schedule (Surridge-David et al. 1987), that there
were high correlations (of the order of 0.6) between the SSS
and analogue scale ratings of mood. This observation led us
to examine the descriptors defining each level of the SSS. It
appeared that these were not clearly equivalent to one
another and included many that were ambiguously related
to sleepiness. Nor was it clear that the scale represents a
single dimension. For example, Dement (1976) has drawn
attention to the difference between sleepiness and physical
tiredness, a distinction which is supported by clinical
observation (Moldofsky 1989).
We decided, therefore, to investigate the psychometric
properties of the SSS and its component items. Specifically,
we focused on two assumptions underlying the scale. The
first is that the descriptors which define any given level of
sleepiness are different but, nonetheless, equivalent ways of
characterizing a particular level of sleepiness. The second is
that sleepiness, as measured by the descriptors given, is an
unidimensional construct.

35
36 A . MacLean et al.

Table 1 Correspondence between the

(1) Feeling active and vital; alert; wide awake. Stanford Sleepiness Scale and the derived
2-1 feel vital. items
%I feel alert.
1 6 1 feel active.
23-1 feel wide awake.
(2) Functioning at a high level; but not at peak; able to concentrate
6 1 am functioning at a high level but not at peak.
10-1 am able to concentrate.
14-1 am functioning at a high level.
(3) Relaxed; awake; not at full alertness; responsive
7-1 am relaxed.
12-1 feel responsive.
2&I am awake.
22-1 am not at full alertness.
(4) A little foggy, not at peak; let down.
3-1 am not at peak.
1 5 - 1 am a little foggy.
17-1 feel let down.
(5) Fogginess, beginning to lost interest in remaining awake; slowed down.
4 - 1 am experiencing fogginess.
13-1 am beginning to lose interest in remaining awake.
21-1 feel slowed down.
(6) Sleepiness; prefer to be lying down; fighting sleep; woozy.
1-1 am fighting sleep.
8-1 am experiencing sleepiness.
19-1 would prefer to be lying down.
24-1 feel woozy.
(7) Almost in reverie, sleep onset soon; lost struggle to remain awake
5 - 1 feel sleep onset will be soon.
11-1 am almost in reverie.
18-1 am losing the struggle t o remain awake.

responses were obtained by the administration of the

METHOD
Each item of the SSS was rewritten as a number of
True/False questions, which resulted in a 24 item
questionnaire (see Table 1). Additional items included the
subjects age and sex; the time at which the questionnaire
was completed; four visual analogue scales rated 0-100
(calm-irritable; happy-sad; energetic-sluggish; relaxed-
tense); the Stanford Sleepiness Scale; and a five-point scale
rating of circadin tendency.
Subjects and procedure
Three hundred and sixty seven undergraduate students at
Queens University volunteered and gave their informed
consent. Their mean age was 21.1 years (s.d. 5.1 years;
range 16 to 56 years), and 52.6% rated themselves as being
definitely or somewhat an evening type compared with
30.6% who described themselves as being definitely or
somewhat a morning type.
They participated in the study by filling out the
questionnaire (described above) on one occasion. After 27
forms were rejected due to missing items or failure to follow
the instructions, a total of 340 (97 male; 243 female)
subjects were included in the analysis. The majority of
questionnaire at the end of a class (175 subjects at 11.00
hours; 123 subjects at 20.00 hours), the remainder of the
responses were obtained individually at times ranging from
09.00 to 23.00 hours. T h e different times of day facilitated a
greater range of SSS scores.
RESULTS
Consistency of scale items
If the adjectives or descriptors that comprise each of the
seven levels on the SSS are internally consistent, then
subjects who rated themselves at a particular SSS level
should also endorse all of the statements generated from
those adjectives or descriptors. Subjects who rated
themselves at, for example, level 3 on the SSS should
endorse items 7, 12, 20 and 22 as True 100% of the
time and all other items 0% of the time. These data
are shown in Fig. 1, where it is clear that this pattern is not
seen. Individuals who rate themselves at level 3 on the SSS
do show high endorsement levels for the four relevant items.
However, they also endorse items 3 and 10 on the
questionnaire. These are associated, respectively, with levels
4 and 2 of the SSS, which are endorsed as often, or more
 Evaluation of Stanford Sleepiness Scale 31

I n.12 2
I oc
8C
6C
40

2c
0
3 n.151 I 4 n.35
100 la

80
E
2
.-
6C
--
0

a,
3
40
s 20
cn
; o
E
m 5 n=27 I 6 0.27
6 100
a,
0
e
80
c
a,
2 60
a,
a
40
20

0
I00
7 n=i I 4 8 12 16 20 24
Figure 1. Proportion of subjects who Item number
respond True to each of the derived items 80
in relation to their score on the Stanford
Sleepiness Scale (SSS). The level of the SSS 60
endorsed by the subjects is shown in the
upper left corner of each part of the figure, 40
the number of the subjects making the
endorsement in the upper right, while the 20
solid bars indicate those items that are
derived from the level of the SSS. Note: 0
only one subject responded to SSS scale 4 8 12 16 20 24
point 7. Item number

often than the relevant items. Similar patterns were also
obtained for subjects rating themselves at levels 1 , 2 , 4 , 5
and 6 of the SSS. Note that the data for level 7 are based on
the responses of only one individual.
Another measure of the consistency of the individual
components of the SSS is the modal level of the SSS which is
selected by individuals who endorse a particular item as
true. Again, if the component items of the SSS are
consistent with the level of the scale from which they are
derived, subjects who endorse, for example, item 19 as
True should all rate themselves at level 6 of the SSS. In
fact, the majority of individuals who endorse item 19 as
True select levels 3 of the SSS. When considering all of the
24 items of the questionnaire, in only six is the modal level
of the SSS selected the one that would be predicted from the
derivation of the item. For ten items it differs by one
position; for four items by two positions; on three items by
three positions; and on one item by four positions.
From neither of these analyses can the component items
of the SSS levels b e judged t o be consistent with the level of
the scale from which they are derived.
Dimensionality of the scale
To explore the dimensionality of the SSS, we carried out a
Principal Components Analysis of the 24 items derived from
the scale descriptors. Two components were extracted,
which were based on the examination of a scree plot (Cattell
38 A . MacLean et al.
Table 2 Principal Components Analysis of
Factor revised Sleep Scale items
Item ~~~

no. 1 2 Item

21 0.743 0.225 I feel slowed down.

22 0.663 0.110 I am not at full alertness.
8 0.634 0.406 I am experiencing sleepiness.
15 0.579 0.307 I am a little foggy.
19 0.558 0.380 1 would prefer to be lying down.
3 0.488 -0.068 I am not at peak.
4 0.421 0.389 I am experiencing fogginess.
6 -0.539 -0.382 I am functioning at a high level but not at peak
2 -0.616 -0.049 I feel vital.
9 -0.709 -0.389 I feel alert.
14 -0.729 -0.271 I am functioning at a high level.
16 -0.740 -0.205 I feel active.
23 -0.770 -0.200 I feel wide awake.
18 0.077 0.762 I am losing the struggle to remain awake.
13 0.232 0.696 I am beginning to lose interest in remaining awake
11 -0.003 0.628 I am almost in reverie.
1 0.320 0.605 I am fighting sleep.
24 0.244 0.590 I feel woozy.
5 0.280 0.513 I feel sleep onset will be soon.
17 0.213 0.507 I feel let down.
12 -0.370 -0.554 I feel responsive.
20 -0.043 -0.612 I am awake.
10 -0.123 -0.693 I am able to concentrate.
7 -0.095 -0.135 I am relaxed.

1966) and they were rotated to a Varimax Criterion: factor Table 3 Pearson product-moment correlations between Visual
loadings greater than 0.320 (representing a minimum of 10 Analogue Scales and Sleepiness Scales
percent of shared variance between the items and the factor) Measure sss Factor 1 Factor 2
were interpreted.
The results of the analysis are shown in Table 2 . The first sss 0.500*** 0.62 1* * *
Age -0.113 -0.055 -0.150
factor, accounting for 24.2 percent of the variance, appears Circadian type 0.113 0.147 0.109
to be an energic or activation factor which is characterized
Calm-Irritable 0.351 * * * 0.269** * 0.321 * * *
at one extreme by the statements I feel slowed down and Energetic-Sluggish 0.668* * * 0.671*** 0.436***
I am not at full alertness and at the other by I feel Happy-Sad 0.36S** * 0.342** 0.291***
active and I feel wide awake. The second factor, which Relaxed-Tense 0.228** 0.219** 0.188*
accounts for 20.6 percent of the variance, seems to be more ~

* = P < 0.05; * * = P < 0.01; *** = P < 0.001. Bonferroni corrected.

related to consciousness, sleepiness and a loss of control
over the ability to remain awake. It is characterized by the
statements I am losing the struggle to remain awake, I
am beginning to lose interest in remaining awake on the
one hand, and on the other hand, by the statements I am
awake. and I am able to concentrate.
the correlations are modest in size and d o not differentiate
substantially between the factors.
There is no systematic relationship between the SSS,
Factor 1 or Factor 2 and age or circadian type.

Analogue scales
To investigate further the interpretation of the two factors,
Pearson product-moment correlations were calculated
between the visual analogue scales, the SSS and the factor
scores of the two factors (Table 3). T h e two largest
correlations, 0.668 and 0.671, respectively, are those
between the Energetic-Sluggish scale us the SSS, and the
Energetic-Sluggish scale us Factor 1. Otherwise, however,
DISCUSSION
The results of the present study indicate that items derived
from the descriptors defining each level of the SSS are not
homogeneous. Furthermore, it was shown that sleepiness, as
defined by these items, is not an unidimensional construct;
two independent dimensions are present. If one considers
the items that load on each component and their
relationship to visual analogue scale ratings, then these can
be identified provisionally as activation and sleepiness.

The multifaceted nature of subjective states related to
sleepiness, and of sleepiness itself, has been recognized by a
number of writers. Although Dement (1976) maintained
that sleepiness was an independent feeling state, he drew
attention to some overlap between sleepiness, physical
tiredness, and depression and lack of initiative. Carskadon
and Dement (1982), in providing a theoretical framework
for the Multiple Sleep Latency Test, distinguished between
physiological and manifest sleepiness o r sleep tendency. The
former is exposed only in the absence of alerting factors,
whilst the latter is influenced by many such factors.
Broughton (1982) has emphasized some of the shortcom-
ings of the SSS. He pointed out that, while the SSS had
been shown to predict performance decrements following
acute sleep deprivation, both under conditions of cumulative
partial sleep deprivation (Herscovitch and Broughton
1981b) and in patients with narcolepsy-cataplexy (Valley
and Broughton 1981) these findings have not been
replicated. Following a comprehensive review of subjective,
behavioural and physiological measures of sleepiness, he
concluded that it seems apparent that the assumption that
excessive sleepiness is a unitary phenomenon varying only in
degree should be abandoned (p. S144).
Horne (1988) has also suggested that more than one type
of sleepiness may exist. H e postulated the existence of two
forms deriving from his proposed division of sleep into core
sleep and optional sleep. H e argues that: we might well
consider looking for two types of sleepiness. One relates to
increased cerebral impairment . . . The other relates to a
build-up of whatever governs optional sleep . . . (p. 57). It
is tempting to speculate that Factor 2 is closely aligned to
optional sleepiness, whereas Factor 1 is aligned to core
sleepiness. If that is the case, then a regimen of partial sleep
deprivation that allows for the integrity of core sleep should
result in altered scores on Factor 2 but not on Factor 1. By
the same token, scores on the former, but not the latter,
should vary in a circadian manner.
While we have tentatively named the components that
emerged from our analysis, we emphasize that these are
only provisional designations. Whether these have any
construct validity is a matter that can only be determined by
appropriate studies; some of which we are presently carrying
out. By so doing, we hope to clarify the interpretation of the
dimensions, to aid in the clarification of the conceptual basis
for the understanding of subjective sleepiness, and to make
some progress towards the construction of a new
Evaluation of Stanford Sleepiness Scale 39
psychometric instrument for the assessment of subjective
sleepiness.
ACKNOWLEDGEMENT
The support of the Medical Research Council of Canada
(Grant No. MA-6937) is gratefully acknowledged.
REFERENCES
Broughton, R. Performance and evoked potential measures of
various states of daytime sleepiness. Sleep, 1982, 5: S135-146.
Carskadon, M. A . and Dement, W. C. The Multiple Sleep Latency
Test: What does it measure? Sleep, 1982, 5: S67-72.
Cattell, R . B. The scree test for the number of factors. Multivariate
Behav. Res., 1966, 1: 245-276.
Dement, W. C . Daytime sleepiness and sleep attacks. In: C .
Guilleminault, W. C. Dement and P. Passouant (Eds)
Narcolepsy. Spectrum Publications, New York 1976: 17-42.
Dement, W. C . , Carskadon, M. A. and Richardson, G. In: C .
Guilleminault and W. C. Dement (Eds) Sleep Apnea Syndromes.
Alan R. Liss, New York 1978: 23-46.
Friedman, J . , Globus, G., Huntley, A , , Mullaney, D . , Naitoh, P.
and Johnson, L. Performance and mood during and after gradual
sleep reduction. Psychophysiol., 1977, 14: 245-250.
Glenville, M. and Broughton, R. Reliability of the Stanford
Sleepiness Scale compared to short duration performance tests
and the Wilkinson Auditory Vigilance Task. in: P. Passouant and
I. Oswald (Eds) Pharmacology of the States of Alertness.
Pergamon Press, Oxford 1979: 235-244.
Herscovitch, J . and Broughton, R. Sensitivity of the Stanford
Sleepiness Scale to the effects of cumulative partial sleep
deprivation and recovery oversleeping. Sleep, 1981a, 4: 83-92.
Herscovitch, J . and Broughton. R. Performance deficits following
short-term partial sleep deprivation and subsequent recovery
oversleeping. Can. J . Psychol., 1981b, 35: 309-322.
Hoddes, E., Dement, W. and Zarcone, V. The development and
use of the Stanford Sleepiness Scale (SSS). Psychophysiol., 1972,
9: 150.
Hoddes, E., Zarcone, V., Smythe, H . , Phillips, R. and Dement,
W. C. Quantification of sleepiness: A new approach.
Psychophysiol., 1973, 10: 431-436.
Horne, J . Why W e Sleep: the functions of sleep in humans and other
mammals. Oxford University Press, Oxford, 1988.
Moldofsky, H. Sleep and fibrositis syndrome. Rheumatic Disease
Clinics of North America, 1989, 15: 91-103.
Surridge-David, M., MacLean, A . W., Coulter, M. and Knowles, J .
B. Mood changes following an acute delay of sleep. Psychiatry
Res., 1987, 22: 149-158.
Valley, V. and Broughton, R. Daytime performance deficits and
physiological vigilance in untreated patients with narcolepsy-
cataplexy compared to controls. Rev. Electroencephalogr.
Neurophysiol. Clin., 1981, 11: 133-139.