Discussion
The results showed that the attention task was per-
formed faster in noise but at lesser accuracy. This
result supports the SATO-hypothesis suggested by
Hockey (1984), stating that working memory in-
creases its speed in noise, but at the expense of in-
creasing errors.
The signi¢cant interactions between noise, heat
and indoor lighting obtained in the present experi-
ment were on the recall of the text (Noise6Heat,
Figure 2 and 4a) and on the free recall of the emo-
tionally toned words (Noise6Light, Figure 3 and
4b). The ¢rst interaction, see Figure 2 and 4a, was
crossover in nature, meaning impaired recall with
heat at the high noise level, but improved recall at
the low noise level. As evident from Figure 4b, this
is not consistent with the assumption of an optimal
performance in the mid-arousal region, but fairly
consistent with noise increasing arousal and mild
heat decreasing it. The second interaction, see
Figure 3 and 4b, was also a crossover interaction
but the ordering of the activation states and the
lack of an in£ection point in Figure 4b, is not con-
sistent with the arousal model's prediction that in-
creased illuminance and noise level would increase
activation.
For the a¡ective states other than HA (and the
correlated AP) in the circumplex model, there were
no indicators of di¡erential main e¡ects of and in-
teractions between noise, heat and lighting.
Further, there were no signs of a curve-linear rela-
tionship between a¡ect states and cognitive perfor-
mance peaking in the middle region of the reported
a¡ect scores.
It can be argued that an ideal test of the inverted
U-hypothesis should have at least three levels each
of the independent variables in order to detect any
in£ection point or peak in performance. Even if
there is no task for which we know the relationship
between arousal and performance, and at what level
arousal performance will peak, we would assume
that for several tasks with di¡erent degrees of di¤-
culty, some of them, if the inverted-U hypothesis is
correct, would show improvement in performance
with slightly lowered activation and impairment
with further lowering of the activation, and an or-
dering of the activation states from combined ex-
perimental conditions (cf. Figures 4a and 4b)
consistent with earlier research. Lacking this im-
provement with decreased activation in the present
study, and ¢nding performance minima, rather than
maxima, in the medium range of activation, does
not lend support to the inverted-U hypothesis and
the arousal model.
Thus, perceived a¡ect states are not likely media-
tors of e¡ects on cognition. In particular the arou-
sal and the inverted U^hypotheses are not suitable
explanatory models, at least not for the levels of
noise, heat and indoor lighting commonly used in
dwellings, schools and o¤ces.
As regards the gender e¡ects, the three-way inter-
action e¡ects between gender, heat and light and
gender, heat and noise on attention are not in line
with the SATO hypothesis suggesting a speed-
accuracy-trade-o¡ e¡ect on working memory and at-
tention. Moreover, compared to men, females