Detecting awareness from autonomic responses 1
DETECTING CONSCIOUS AWARENESS FROM INVOLUNTARY AUTONOMIC
RESPONSES
Ryan B. Scott D.Phil. 1, 4
Ludovico Minati MSc. 2, 5
Zoltan Dienes D.Phil. 1, 4
Hugo D. Critchley D.Phil. 2, 4
Anil K. Seth D.Phil. 3, 4
1
2
School of Psychology, University of Sussex
Department of Psychiatry, Brighton & Sussex Medical School (BSMS)
3
4
5
School of Informatics, University of Sussex
Sackler Centre for Consciousness Science, University of Sussex
Scientific Department, Neuroradiology and Clinical Neurophysiology
Units, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milano,
Italy.
Corresponding author:
Dr Ryan Scott
Pevensey Building
School of Psychology
University of Sussex
Falmer
BN1 9QH
Tel: +44 (0)1273 876650
Fax: +44 (0)1273 678058
Email: r.b.scott@sussex.ac.uk
Detecting awareness from autonomic responses 2
Abstract
Can conscious awareness be ascertained from physiological responses alone? We
evaluate a novel learning-based procedure permitting detection of conscious awareness
without reliance on language comprehension or behavioural responses. The method exploits a
situation whereby only consciously detected violations of an expectation alter skin
conductance responses (SCRs). 30 participants listened to sequences of piano notes that,
without their being told, predicted a pleasant fanfare or an aversive noise according to an
abstract rule. Stimuli were presented without distraction (attended), or while distracted by a
visual task to remove awareness of the rule (unattended). A test-phase included occasional
violations of the rule. Only participants attending the sounds reported awareness of violations
and only they showed significantly greater SCR for noise occurring in violation, versus
accordance, with the rule. Our results establish theoretically significant dissociations between
conscious and unconscious processing and furnish new opportunities for clinical assessment
of residual consciousness in patient populations.
Keywords:
Consciousness
Conscious awareness
Disorders of Consciousness
Persistent Vegetative State
Minimally Conscious State
Locked-In Syndrome
Unconscious knowledge
Detecting awareness from autonomic responses 3
1. Introduction
An important challenge for cognitive psychology, neuroscience, and clinical
neurology is to determine whether conscious awareness can be detected through nonbehavioural responses. From a theoretical perspective, any method allowing the conscious
status of knowledge to be assessed without the confounding influence of explicit subjective
reports is of considerable value. More urgently, the assessment of consciousness in brain
injured patients is central to the differential diagnosis of vegetative state, minimally
conscious state, and locked-in syndrome. Accurate diagnosis is needed to inform prognosis
and clinical management (Jennett, 2002). Current clinical practice relies heavily on
behavioural evaluation and is therefore limited where volitional motor responses or language
comprehension are compromised (Majerus, Gill-Thwaites, Andrews, & Laureys, 2005).
These limitations contribute to rates of misdiagnosis, estimated to be 37-43% in patients
diagnosed as vegetative state (Andrews, Murphy, Munday, & Littlewood, 1996; Childs,
Mercer, & Childs, 1993; Schnakers et al., 2009).
A variety of methods have been developed to aid evaluation of behaviourally
unresponsive patients. Electrophysiological and nuclear medicine techniques have established
power for predicting negative outcome (Carter & Butt, 2005; Daltrozzo, Wioland, Mutschler,
& Kotchoubey, 2007). In contrast, detecting the presence of residual consciousness as a
means to predict positive outcome, remains a substantial challenge.
Cognitive theories and empirical research have motivated a number of candidate
methods. One approach draws on an apparent awareness-dependence for trace versus delay
conditioning. In trace conditioning the CS ends prior to the US, thus requiring a memory
‘trace’ for an association to be made, while in delay conditioning the CS and US overlap.
Several studies have found that trace but not delay conditioning requires awareness of the
stimuli and of the associative relationship (Knight, Nguyen, & Bandettini, 2006; Lovibond &
Detecting awareness from autonomic responses 4
Shanks, 2002), though others have failed to observe this limitation to unconscious learning
(Destrebecqz et al., 2010; Fu, Fu, & Dienes, 2008). Bekinschtein et al., (2009) employed
trace conditioning of the eye-blink response evaluating anticipatory electromyographical
responses as an indicator of learning. While they observed that the degree of learning was a
good indicator of recovery in patients, the method failed to provide a clean separation of
conscious and unconscious control subjects; some conscious controls failed to learn and one
unconscious control showed marginal learning while under general anaesthetic.
A related approach exploits cognitive event-related potentials (ERPs) to detect
awareness of violations in temporal regularities. Bekinschtein et al. (2009) devised a
paradigm with both local and global violations of an auditory regularity such that detecting
the global differences required maintaining the perceptual representation over some seconds,
which was held to require consciousness. Consistent with their predictions only the global
violations generated a P300 ERP complex previously associated with conscious access
(Sergent, Baillet, & Dehaene, 2005), and this effect was only present in participants reporting
awareness of the global structure. However, the effect was only reliably observed in
participants instructed to attend to the global regularity. The method is consequently
dependent on verbal comprehension which is known to be impaired in a significant
proportion of stroke and traumatic brain injury patients (Eisenberg et al., 1990; Inatomi et al.,
2008).
Finally, an alternative approach employs functional neuroimaging to index distinct
patterns of regional brain activity associated with the content of intentional visual imagery.
This method has provided evidence for conscious awareness in patients otherwise fulfilling
the criteria for vegetative state (Monti et al., 2010; Owen et al., 2006). However, in addition
to requiring brain-imaging equipment, the method is again reliant on verbal comprehension;
patients must change their mental imagery in accordance with verbal instructions.
Detecting awareness from autonomic responses 5
Despite these advances, there remains a need for an accurate method to determine conscious
awareness, that is not reliant on motor responses or language comprehension, and which
preferably can be applied at the bedside. To address this challenge we devised a novel
approach which we term the Learned Aversive Contingency (LAC) procedure. The method
exploits the skin conductance response (SCR) to index learning of a predictive relationship.
Specifically, sequences of piano notes predicted either a pleasant fanfare or aversive white
noise. Patients diagnosed as persistent vegetative state (PVS) are known to exhibit significant
SCRs in response to white noise (Hildebrandt, Zieger, Engel, Fritz, & Bussmann, 1998;
Keller, Hulsdunk, & Muller, 2007), making this a suitable clinical measure. The predictive
sequences were designed to minimise the likelihood of unconscious learning. In common
with previous attempts, a delay was inserted between the predictive sequence and the
aversive or non-aversive stimulus. However, based on evidence that the proportion of
unconscious learning reduces with longer delays (e.g. Gustav Kuhn & Dienes, 2005; G. Kuhn
& Dienes, 2006) we inserted an extended pause of between 1 and 2 seconds during training
and always 2 seconds during testing - four times that commonly used in trace conditioning. In
addition, the predictive rule was abstract in nature; the pattern of similarity between the notes
predicted the outcome but the pitch of the notes was different in every test-sequence. Implicit
learning of this type of consistency, known as repetition structure, can be observed in
paradigms such as artificial grammar learning (AGL), where its influence is mediated by
feelings of familiarity (Scott & Dienes, 2008, 2010a; but not by fluency, Scott & Dienes,
2010b). However, the learning context in paradigms such as AGL is substantially different to
that employed in the current study. Most notably, in AGL all elements of the stimuli are
typically presented either simultaneously as a single visual unit, or as an uninterrupted
stream. In the present auditory paradigm we aimed to reduce implicit learning by a
Detecting awareness from autonomic responses 6
combination of the 2-second delay preceding the final element of each sequence and the use
of a unique pitch on each trial.
To measure conscious expectation, we exploited differences in SCR in response to the
noise when its occurrence was predictable versus unpredictable. SCR magnitudes evoked by
an aversive stimulus are greater when the stimulus is unanticipated (Ohman, 1971). Learning
was therefore assessed by comparing SCR in test trials where the rule either applied as before
or was violated such that the aversive noise was unexpected. Critically, we sought to devise a
training sequence that made the rule sufficiently salient that individuals attending the sounds
would detect it without instruction, thus avoiding dependence on verbal comprehension. We
manipulated attention as the means to simulate patients who are either consciously aware or
unaware of the auditory structure, thus permitting us to test the extent to which learning is
dependent on awareness. While opinions differ as to whether consciousness and attention are
doubly dissociable processes (Bussche, Hughes, Humbeeck, & Reynvoet, 2010; Koch &
Tsuchiya, 2007), we merely exploited their uncontroversial possible association, and verified
the effectiveness of the attention manipulation on conscious awareness of the rule using
subjective reports.
We evaluate the LAC procedure by contrasting reported awareness and SCR
measurements in participants exposed to the sound sequences without distraction with those
of participants engaged in a visual task removing attention from the sounds. Removing
awareness in this way has important advantages over alternative approaches, such as the use
of anaesthesia (cf. Bekinschtein, Shalom et al., 2009), because unlike those approaches
attentional manipulation should not compromise other brain functions possibly preserved in
patients.
Detecting awareness from autonomic responses 7
2. Material and methods
2.1 Participants
Thirty students (11 male, 19 female; age M = 23, SD = 4.1 years) participated in
exchange for course credits or £5. All participants were naive to the experimental hypothesis
and were randomly assigned to experimental condition (attended or unattended).1
2.2 Materials
Sound sequences included 40 training trials and 65 test trials. Each trial consisted of
three piano notes followed by a pause (initially 1 s, increasing to 2 s) and either a pleasant
fanfare or a burst of white noise (100 dB, 1 s). Two trial types were presented during
training: 1) non-aversive trials, consisting of three identical notes followed by the fanfare,
and 2) aversive trials, where the third note was substantially different from the previous two
and followed by white noise. Training included 15 aversive trials and 25 non-aversive trials
arranged to facilitate uninstructed learning of a simple rule: Three identical notes (of any
pitch) predicts the pleasant fanfare, whereas a different third note predicts the noise burst.
The test phase consisted of eight blocks of eight trials with the first block preceded by
one additional aversive trial. This first aversive trial was included solely to reacclimatise
participants to the noise burst after the five minutes of silence following training; this trial
was therefore not included in the analyses. Each of the eight blocks contained six nonaversive trials, one predictable aversive trial, and one unpredictable aversive trial, in pseudorandom order. In predictable aversive trials, the noise burst was preceded by a sequence in
which the third note was different, as in training. In unpredictable aversive trials, the noise
burst occurred after three identical notes, violating the rule. Learning would be apparent as
1
Twenty four participants were randomly assigned in equal numbers to the two conditions. To replace those not
showing sufficient sympathetic response for learning assessment (see results), an additional six were recruited
and randomly assigned in appropriate proportions.
Detecting awareness from autonomic responses 8
larger magnitude SCRs for unpredictable compared with predictable trials (Figure 1). The
pause between piano notes and fanfare or white noise was 2 s in all test trials. Full details of
training and test sequences are given in Tables S1 and S2 of the supplementary material.
Non-aversive
trial
Unpredictable
aversive trial
NOISE
Predictable
aversive trial
NOISE
SCR - attended
condition
Difference
indicative of
rule learning
SCR - unattended
condition
Figure 1. Trial types and predicted SCR. In the attended condition, larger magnitude responses are
expected for unpredictable as compared to predictable trials, indicating rule learning. In the
unattended condition, this difference is expected to be abolished.
2.3 Procedure
Participants were seated at a computer and equipped with headphones. Skin
conductance was recorded from electrodes attached to the index and middle fingers of their
left hand. Training sequences were followed by a five minute rest before the test sequences.
The attended condition aimed to simulate conscious patients without motor volition or
language comprehension. These participants were therefore asked to imagine they were
hospital patients, unable to move or understand language. Care was taken not to provide any
verbal or other instruction. The unattended condition aimed to simulate patients lacking
conscious awareness of the stimuli. These participants were therefore asked to perform a
visual discrimination task as accurately as possible while ignoring the sounds. The visual
Detecting awareness from autonomic responses 9
task involved using eye-movements to indicate the position and orientation of a line
appearing on screen (see Figure S1 in the supplementary material).
3. Results
Skin conductance was measured with a sensitivity of 0-100 µS, the signal sampled at
500 Sa/s 16-bit, low-pass filtered at 5 Hz, and detrended with a fifteen degree polynomial.
For each test-trial, the amplitude of SCR evoked by the fanfare or noise burst was determined
as the difference between the maximum in the 0-8 s post-stimulus interval and the average in
the 1-3 s pre-stimulus interval.
Statistical analysis was performed individually for each participant. First, SCRs with
amplitude beyond three standard deviations from the mean for each trial type were excluded
as outliers; there were 36 outliers in total, no more than two for any participant, and all were
for non-aversive trials.2 Second, a two-sample t-test was performed to verify whether SCR
amplitude was greater for noise bursts than fanfares; since the absence of this difference
would indicate insufficient sympathetic response to use SCR as a measure of learning, such
participants were excluded from further assessment. This represents a practical constraint but
also ensures that in clinical application compromised autonomic responsivity would not result
in a false negative assessment. In the remaining participants, a paired one-tailed t-test
assessed whether SCR magnitudes were greater for the eight unpredictable compared with the
eight predictable aversive trials.
Upon completion of the experiment, each participant in the attended condition reported
learning the rule during the training phase and realizing that it was occasionally violated
2
This pattern is to be expected given that there were more non-aversive than aversive trials (48 vs. 16) and that
the magnitude of SCRs elicited is smaller for non-aversive than aversive trials. As such, any compromised
response, for example resulting from the participant coughing, will have been more likely to occur during a nonaversive trial and would also be more likely to result in an SCR magnitude outside the normal range for nonaversive than aversive trials.
Detecting awareness from autonomic responses 10
during the test phase. No participant in the unattended condition reported awareness of the
relationship between the piano notes and subsequent sounds; as intended, the visual
distraction task eliminated conscious awareness of the auditory structure. Six participants
(one attended condition, five unattended condition) failed to show significantly greater SCR
magnitudes to noise bursts than fanfares and were hence excluded from the subsequent
learning assessment.
40
4
*
Group Means
* Participant mean greater than zero p < .05
30
3
+ Participant mean greater than zero p < .10
MeanSCR
GSRDifference
Difference(μS)
Mean
*
*
20
2
*
*
10
1
+
*
*
*
*
*
*
00
-10
-1
-20
-2
-30
-3
Attended Condition
Unattended Condition
(7 female, 5 male)
(7 female, 5 male)
Figure 2. Mean difference in SCR magnitude for unpredictable minus predictable aversive trials,
shown for each participant grouped by condition. Error bars indicate standard error of the mean.
For the remaining twenty four participants (12 per condition) the differences in SCR
magnitudes for unpredictable minus predictable aversive trials are shown in Figure 2. For the
Detecting awareness from autonomic responses 11
attended condition, SCR magnitudes were larger for unpredictable as compared to predictable
aversive trials (M = 20.9, SD = 10.8 µS vs. M = 11.2, SD = 7.8 µS). This effect was
significant at the individual level for eleven participants (p < .05) and marginal for one (p =
.071). In the unattended condition, SCR magnitudes were not larger for unpredictable
compared with predictable trials (M = 4.9, SD = 6.2 µS vs. M = 6.6, SD = 8.3 µS); the effect
was not significant at the individual level for any participant (all p > .13; 95% CI of the
difference -3.4, +1.0 µS, ruling out population differences even remote from the attended
mean difference of 10.9 µS).
It was possible that the visual task could have influenced SCR magnitudes in some
way other than by the intended influence on awareness of the auditory structure. To test for
this we compared the magnitude of the SCRs for each type of test trial, between attended and
unattended conditions, see Figure 3. The magnitudes did not differ between groups for nonaversive trials, t(22) = .94, p = .356, or predictable aversive trials, t(22) = 1.03, p = .314, but
did so for unpredictable aversive trials, t(22) = 4.79, p < .001, where, as predicted based on
the violation of expectations, SCR magnitudes were significantly greater for the attended
condition. There is hence no evidence that the visual task affected SCRs by a means other
than the intended influence on awareness.
Detecting awareness from autonomic responses 12
3.0
Unattended
Attended
2.5
Mean SCR (μS)
2.0
1.5
1.0
0.5
0.0
Non-aversive
Predictable
aversive
Unpredictable
aversive
Figure 3. Mean SCR magnitudes for each type of test trial contrasted between attended and
unattended conditions. Error bars indicate standard error of the mean.
The results demonstrate that only participants who had conscious awareness of the
rule showed a significantly greater SCR magnitude when it was violated. Participants in the
unattended condition, all of whom reported being unaware of the rule, showed equivalent
SCR magnitudes after white noise irrespective of whether it occurred in accordance or
violation of the rule. It is not possible to estimate the extent to which the observed absence of
implicit learning was due to any specific feature of the paradigm. However, the results are
consistent with both the inclusion of an extended pause prior to the predicted stimulus and the
use of different pitches on each trial being effective in reducing unconscious learning.
Detecting awareness from autonomic responses 13
4. Discussion
The results confirm that 1) the LAC procedure enabled reliable learning of an abstract
relation in the absence of any instruction, 2) conscious awareness of the relation was
important for learning to occur, and 3) learning was detectable from autonomic responses.
The procedure thus provides a means to detect conscious awareness based solely on an
autonomic response and avoids reliance on language comprehension or motor responses.
The technical simplicity of the LAC procedure ensures that it could be easily
implemented at the bedside, inviting consideration of its clinical potential. Our results suggest
the method would have a high diagnostic specificity; i.e., it would provide strong evidence for
the presence of conscious awareness. This is supported by the fact that removing awareness
of the contingency by manipulating attention was sufficient to eliminate learning despite
cognitive function being otherwise intact. The procedure should also provide higher
sensitivity than existing measures because it does not require language comprehension or
motor responses. Potential limitations include inadequate cognitive or attentional capacity,
which may occur in patients with extensive cortical damage, and compromised autonomic
responsivity impairing SCR. While SCRs evoked by white noise are observed in PVS
patients (Hildebrandt, Zieger, Engel, Fritz, & Bussmann, 1998; Keller, Hulsdunk, & Muller,
2007) the procedure also embeds a test for these responses, thus allowing identification of
patients for whom assessment would be inconclusive. Relevant to both specificity and
sensitivity, a recent study found that coma patients have preserved SCR to emotionally salient
stimuli (Daltrozzo et al., 2010). The presence of SCR in presumably unconscious patients
emphasises the robustness, and hence suitability, of this response, while also highlighting the
need to test for learning of an abstract relation to provide specificity.
The absence of a reliable SCR to the white noise observed for five participants in the
non-attending condition may in part be due to these participants being required to actively
Detecting awareness from autonomic responses 14
focus their attention on the visual task. This reduced responsivity would not be expected in
conscious patients as it is unlikely they would deliberately focus their attention on something
other than the sound sequences.
The term consciousness has several meanings (Rosenthal, 2005). For example, there
is a distinction between creature consciousness and mental state consciousness. Our method
bears only on the latter. That is, our aim is to answer the question: Does the person being
assessed have any conscious mental states? By virtue of the empirical reliance that we have
demonstrated, namely that under the conditions of our test conscious knowledge is required
for learning to be apparent, learning on our test can be taken as evidence for the existence of
conscious knowledge. A negative result allows no conclusion as to whether or not the person
has any conscious mental states.
During development of our procedure a variety of experimental factors were found to
influence the reliability with which attending participants acquired the rule. Multiple
revisions of the training sequences were necessary to overcome an apparent contingency
blindness, whereby seemingly obvious relations were missed in the absence of direct
instruction. This phenomenon has parallels with other surprising examples of cognitive
opacity including change blindness and inattentional blindness (Simons & Chabris, 1999;
Simons & Rensink, 2005) and is it itself worthy of further research. For example, such
research could investigate more generally how dissociations between conscious and
unconscious processing depend on task instructions and participant expectations, as well as
on the details of the stimuli. A summary of all factors observed to influence learning is
provided in the supplementary material available online.
Detecting awareness from autonomic responses 15
5. Conclusion
We have demonstrated how learning of an aversive contingency, assessed by an
autonomic response, can be employed to reliably identify conscious awareness independently
of language comprehension or motor volition. The LAC procedure advances our
understanding of dissociations between conscious and unconscious processing in the absence
of explicit instruction. It also has the potential to substantially enhance the clinical assessment
of conscious awareness in brain injured patients.
6. Acknowledgements
The authors are grateful to Paolo Cortellazzi, MD, for insightful feedback on an
earlier version of the manuscript. RS and ZD were supported by an ESRC Project Grant. LM
was supported by a Wellcome Trust Programme Grant to HDC. Participant funding was
provided by the Dr Mortimer and Theresa Sackler Foundation.
Detecting awareness from autonomic responses 16
7. References
Andrews, K., Murphy, L., Munday, R., & Littlewood, C. (1996). Misdiagnosis of the
vegetative state: Retrospective study in a rehabilitation unit. British Medical Journal,
313(7048), 13-16.
Bekinschtein, T. A., Dehaene, S., Rohaut, B., Tadel, F. O., Cohen, L., & Naccache, L.
(2009). Neural signature of the conscious processing of auditory regularities.
Proceedings of the National Academy of Sciences of the United States of America,
106(5), 1672-1677.
Bekinschtein, T. A., Shalom, D. E., Forcato, C., Herrera, M., Coleman, M. R., Manes, F. F.,
et al. (2009). Classical conditioning in the vegetative and minimally conscious state.
Nature Neuroscience, 12(10), 1343-U1176.
Bussche, E. V. d., Hughes, G., Humbeeck, N. V., & Reynvoet, B. (2010). The relation
between consciousness and attention: An empirical study using the priming paradigm.
Consciousness and Cognition, 19(1), 86-97.
Carter, B. G., & Butt, W. (2005). Are somatosensory evoked potentials the best predictor of
outcome after severe brain injury? A systematic review. Intensive Care Medicine,
31(6), 765-775.
Childs, N. L., Mercer, W. N., & Childs, H. W. (1993). Accuracy of Diagnosis of Persistent
Vegetative State. Neurology, 43(8), 1465-1467.
Daltrozzo, J., Wioland, N., Mutschler, V., & Kotchoubey, B. (2007). Predicting coma and
other low responsive patients outcome using event-related brain potentials: A metaanalysis. Clinical Neurophysiology, 118(3), 606-614.
Daltrozzo, J., Wioland, N., Mutschler, V., Lutun, P., Calon, B., Meyer, A., et al. (2010).
Emotional electrodermal response in coma and other low-responsive patients.
Neuroscience Letters, 475(1), 44-47.
Detecting awareness from autonomic responses 17
Destrebecqz, A., Perruchet, P., Cleeremans, A., Laureys, S., Maquet, P., & Peigneux, P.
(2010). The influence of temporal factors on automatic priming and conscious
expectancy in a simple reaction time task. Quarterly Journal of Experimental
Psychology, 63(2), 291-309.
Eisenberg, H. M., Gary, H. E., Aldrich, E. F., Saydjari, C., Turner, B., Foulkes, M. A., et al.
(1990). Initial CT Findings in 753 Patients with Severe Head-Injury - A Report from
the NIH Traumatic Coma Data-Bank. Journal of Neurosurgery, 73(5), 688-698.
Fu, Q. F., Fu, X. L., & Dienes, Z. (2008). Implicit sequence learning and conscious
awareness. Consciousness and Cognition, 17(1), 185-202.
Hildebrandt, H., Zieger, A., Engel, A., Fritz, K. W., & Bussmann, B. (1998). Differentiation
of autonomic nervous activity in different stages of coma displayed by power
spectrum analysis of heart rate variability. European Archives of Psychiatry and
Clinical Neuroscience, 248(1), 46-52.
Inatomi, Y., Yonehara, T., Omiya, S., Hashimoto, Y., Hirano, T., & Uchino, M. (2008).
Aphasia during the acute phase in ischemic stroke. Cerebrovascular Diseases, 25(4),
316-323.
Jennett, B. (2002). The vegetative state: medical aspects, ethical and legal dilemmas.
Cambridge, England: Cambridge University Press.
Keller, I., Hulsdunk, A., & Muller, F. (2007). The influence of acoustic and tactile
stimulation on vegetative parameters vegetative state and EEG in persistent.
Functional Neurology, 22(3), 159-163.
Knight, D. C., Nguyen, H. T., & Bandettini, P. A. (2006). The role of awareness in delay and
trace fear conditioning in humans. Cognitive, Affective, & Behavioral Neuroscience,
6(2), 157-162.
Detecting awareness from autonomic responses 18
Koch, C., & Tsuchiya, N. (2007). Attention and consciousness: two distinct brain processes.
Trends in Cognitive Sciences, 11(1), 16-22.
Kuhn, G., & Dienes, Z. (2005). Implicit Learning of Nonlocal Musical Rules: Implicitly
Learning More Than Chunks. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 31(6), 1417-1432.
Kuhn, G., & Dienes, Z. (2006). Differences in the types of musical regularity learnt in
incidental- and intentional-learning conditions. Quarterly Journal of Experimental
Psychology, 59(10), 1725-1744.
Lovibond, P. F., & Shanks, D. R. (2002). The role of awareness in Pavlovian conditioning:
Empirical evidence and theoretical implications. Journal of Experimental Psychology:
Animal Behavior Processes, 28(1), 3-26.
Majerus, S., Gill-Thwaites, H., Andrews, K., & Laureys, S. (2005). Behavioral evaluation of
consciousness in severe brain damage. Boundaries of Consciousness: Neurobiology
and Neuropathology, 150, 397-413.
Monti, M. M., Vanhaudenhuyse, A., Coleman, M. R., Boly, M., Pickard, J. D., Tshibanda, L.,
et al. (2010). Willful Modulation of Brain Activity in Disorders of Consciousness.
New England Journal of Medicine, 10.1056/NEJMoa0905370. .
Ohman, A. (1971). Interaction between instruction-induced expectancy and strength of
unconditioned stimulus in GSR conditioning. Journal of Experimental Psychology,
88(3), 384-390.
Owen, A. M., Coleman, M. R., Boly, M., Davis, M. H., Laureys, S., & Pickard, J. D. (2006).
Detecting awareness in the vegetative state. Science, 313(5792), 1402-1402.
Rosenthal, D. M. (2005). Consciousness and Mind. Oxford: Clarendon Press.
Detecting awareness from autonomic responses 19
Schnakers, C., Vanhaudenhuyse, A., Giacino, J., Ventura, M., Boly, M., Majerus, S., et al.
(2009). Diagnostic accuracy of the vegetative and minimally conscious state: Clinical
consensus versus standardized neurobehavioral assessment. BMC Neurology, 9.
Scott, R. B., & Dienes, Z. (2008). The conscious, the unconscious, and familiarity. Journal of
Experimental Psychology-Learning Memory and Cognition, 34(5), 1264-1288.
Scott, R. B., & Dienes, Z. (2010a). Knowledge applied to new domains: The unconscious
succeeds where the conscious fails. Consciousness and Cognition, 19, 391-398.
Scott, R. B., & Dienes, Z. (2010b). Fluency does not express implicit knowledge of artificial
grammars. Cognition, 114(3), 372-388.
Sergent, C., Baillet, S., & Dehaene, S. (2005). Timing of the brain events underlying access
to consciousness during the attentional blink. Nature Neuroscience, 8(10), 1391-1400.
Simons, D. J., & Chabris, C. F. (1999). Gorillas in our midst: Sustained inattentional
blindness for dynamic events. Perception, 28(9), 1059-1074.
Simons, D. J., & Rensink, R. A. (2005). Change blindness: Past, present, and future. Trends
in Cognitive Sciences, 9(1), 16-20.
Detecting awareness from autonomic responses 20
SUPPLEMENTAL MATERIALS
Detecting awareness from autonomic responses 21
Table S1. Training trial sequences.
Inter-trial
Trial No.
Note 1
Pause
Note 2
Pause
Note 3
Pause
Stimulus
pause
1
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
2
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
3
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
4
C3
100ms
C3
100ms
D#1
1000ms
white noise
2500ms
5
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
6
C3
100ms
C3
100ms
D#1
1000ms
white noise
2500ms
7
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
8
C3
100ms
C3
100ms
C3
1000ms
fanfare
2500ms
9
C3
100ms
C3
100ms
D#1
1000ms
white noise
2500ms
10
C3
100ms
C3
100ms
C3
1000ms
fanfare
10000ms
11
D4
100ms
D4
100ms
D4
1000ms
fanfare
2500ms
12
D4
100ms
D4
100ms
D4
1000ms
fanfare
2500ms
13
D4
100ms
D4
100ms
F#1
1000ms
white noise
2500ms
14
D4
100ms
D4
100ms
D4
1000ms
fanfare
2500ms
15
D4
100ms
D4
100ms
F#1
1000ms
white noise
10000ms
16
F#3
100ms
F#3
100ms
F#3
1000ms
fanfare
2500ms
17
F#3
100ms
F#3
100ms
G2
1000ms
white noise
2500ms
18
F#3
100ms
F#3
100ms
F#3
1000ms
fanfare
2500ms
19
F#3
100ms
F#3
100ms
G2
1000ms
white noise
2500ms
Detecting awareness from autonomic responses 22
20
F#3
100ms
F#3
100ms
F#3
1000ms
fanfare
10000ms
21
B3
100ms
B3
100ms
G#2
1050ms
white noise
2675ms
22
B3
100ms
B3
100ms
B3
1100ms
fanfare
2850ms
23
B3
100ms
B3
100ms
G#2
1150ms
white noise
3025ms
24
B3
100ms
B3
100ms
G#2
1200ms
white noise
3200ms
25
B3
100ms
B3
100ms
B3
1250ms
fanfare
10000ms
26
G5
100ms
G5
100ms
G5
1300ms
fanfare
3550ms
27
G5
100ms
G5
100ms
C#3
1350ms
white noise
3725ms
28
G5
100ms
G5
100ms
C#3
1400ms
white noise
3900ms
29
G5
100ms
G5
100ms
G5
1450ms
fanfare
4075ms
30
G5
100ms
G5
100ms
G5
1500ms
fanfare
10000ms
31
G2
100ms
G2
100ms
G2
1550ms
fanfare
4425ms
32
F5
100ms
F5
100ms
F5
1600ms
fanfare
4600ms
33
C2
100ms
C2
100ms
C2
1650ms
fanfare
4775ms
34
C#3
100ms
C#3
100ms
D1
1700ms
white noise
4950ms
35
F#1
100ms
F#1
100ms
F#1
1750ms
fanfare
10000ms
36
A3
100ms
A3
100ms
A3
1800ms
fanfare
5300ms
37
E5
100ms
E5
100ms
E5
1850ms
fanfare
5475ms
38
E5
100ms
E5
100ms
C2
1900ms
white noise
5650ms
39
B4
100ms
B4
100ms
B4
1950ms
fanfare
5825ms
40
D#2
100ms
D#2
100ms
C1
2000ms
white noise
-
Detecting awareness from autonomic responses 23
Note. The piano notes are written with the octave number. For example, D#2 refers to the
note ‘D sharp’ played in the second octave. The duration of the training phase was
approximately 5 minutes. The piano notes and the stimuli (fanfare or white noise) were
played to different ears; in the first 20 trials the piano notes were played to the left ear and the
stimuli to the right ear, for the remaining 20 trials this was reversed.
Detecting awareness from autonomic responses 24
Table S2. Test trial sequences.
InterBlock
Trial
Note
Note
Pause
No.
No.
1
Note
Pause
2
Type of
Pause
Stimulus
trial
3
aversive trial
pause
-
1
D#2
100ms
D#2
100ms
C1
2000ms
white noise
14000ms
Block 1
2
D#4
100ms
D#4
100ms
D#4
2000ms
fanfare
14000ms
3
D#5
100ms
D#5
100ms
D#5
2000ms
fanfare
14000ms
4
G#3
100ms
G#3
100ms
G#3
2000ms
fanfare
14000ms
5
A#4
100ms
A#4
100ms
D2
2000ms
white noise
14000ms
6
C2
100ms
C2
100ms
C2
2000ms
fanfare
14000ms
7
B5
100ms
B5
100ms
B5
2000ms
fanfare
14000ms
8
D#1
100ms
D#1
100ms
D#1
2000ms
white noise
14000ms
9
B3
100ms
B3
100ms
B3
2000ms
fanfare
14000ms
10
F4
100ms
F4
100ms
F4
2000ms
fanfare
14000ms
11
D#5
100ms
D#5
100ms
G3
2000ms
white noise
14000ms
12
F#5
100ms
F#5
100ms
F#5
2000ms
fanfare
14000ms
13
A#2
100ms
A#2
100ms
A#2
2000ms
fanfare
14000ms
14
F#2
100ms
F#2
100ms
F#2
2000ms
fanfare
14000ms
15
F#4
100ms
F#4
100ms
F#4
2000ms
white noise
14000ms
16
C#4
100ms
C#4
100ms
C#4
2000ms
fanfare
14000ms
17
G3
100ms
G3
100ms
G3
2000ms
fanfare
14000ms
18
D#2
100ms
D#2
100ms
D#2
2000ms
fanfare
14000ms
19
G1
100ms
G1
100ms
G1
2000ms
fanfare
14000ms
Block 2
Block 3
Predictable
Predictable
Unpredictable
Predictable
Unpredictable
Detecting awareness from autonomic responses 25
Block 4
Block 5
20
F5
100ms
F5
100ms
F5
2000ms
fanfare
14000ms
21
E1
100ms
E1
100ms
E1
2000ms
white noise
14000ms
22
C#1
100ms
C#1
100ms
C#1
2000ms
fanfare
14000ms
23
A3
100ms
A3
100ms
A3
2000ms
fanfare
14000ms
24
F3
100ms
F3
100ms
F3
2000ms
fanfare
14000ms
25
F4
100ms
F4
100ms
B2
2000ms
white noise
14000ms
26
C#2
100ms
C#2
100ms
C#2
2000ms
fanfare
14000ms
27
C#3
100ms
C#3
100ms
C#3
2000ms
fanfare
14000ms
28
A4
100ms
A4
100ms
A4
2000ms
fanfare
14000ms
29
B1
100ms
B1
100ms
B1
2000ms
fanfare
14000ms
30
A3
100ms
A3
100ms
G1
2000ms
white noise
14000ms
31
E3
100ms
E3
100ms
E3
2000ms
fanfare
14000ms
32
D3
100ms
D3
100ms
D3
2000ms
fanfare
14000ms
33
B4
100ms
B4
100ms
B4
2000ms
white noise
14000ms
34
A1
100ms
A1
100ms
A1
2000ms
fanfare
14000ms
35
G#5
100ms
G#5
100ms
G#5
2000ms
fanfare
14000ms
36
D2
100ms
D2
100ms
D2
2000ms
fanfare
14000ms
37
G#4
100ms
G#4
100ms
G#4
2000ms
white noise
14000ms
38
A2
100ms
A2
100ms
A2
2000ms
fanfare
14000ms
39
C#5
100ms
C#5
100ms
C#5
2000ms
fanfare
14000ms
40
G#2
100ms
G#2
100ms
E1
2000ms
white noise
14000ms
41
C5
100ms
C5
100ms
C5
2000ms
fanfare
14000ms
Unpredictable
Predictable
Predictable
Unpredictable
Unpredictable
Predictable
Detecting awareness from autonomic responses 26
Block 6
Block 7
Block 8
42
C1
100ms
C1
100ms
C1
2000ms
fanfare
14000ms
43
G4
100ms
G4
100ms
G4
2000ms
fanfare
14000ms
44
E2
100ms
E2
100ms
E2
2000ms
fanfare
14000ms
45
A#4
100ms
A#4
100ms
A#4
2000ms
white noise
14000ms
46
G#1
100ms
G#1
100ms
G#1
2000ms
fanfare
14000ms
47
G#2
100ms
G#2
100ms
G#2
2000ms
fanfare
14000ms
48
B2
100ms
B2
100ms
B2
2000ms
fanfare
14000ms
49
E4
100ms
E4
100ms
A1
2000ms
white noise
14000ms
50
A5
100ms
A5
100ms
A5
2000ms
fanfare
14000ms
51
D1
100ms
D1
100ms
D1
2000ms
fanfare
14000ms
52
A#1
100ms
A#1
100ms
A#1
2000ms
white noise
14000ms
53
F1
100ms
F1
100ms
F1
2000ms
fanfare
14000ms
54
A#3
100ms
A#3
100ms
A#3
2000ms
fanfare
14000ms
55
D#3
100ms
D#3
100ms
D#3
2000ms
fanfare
14000ms
56
G4
100ms
G4
100ms
A#2
2000ms
white noise
14000ms
57
D5
100ms
D5
100ms
D5
2000ms
fanfare
14000ms
58
F#1
100ms
F#1
100ms
F#1
2000ms
fanfare
14000ms
59
B2
100ms
B2
100ms
C#1
2000ms
white noise
14000ms
60
C4
100ms
C4
100ms
C4
2000ms
fanfare
14000ms
61
A#5
100ms
A#5
100ms
A#5
2000ms
fanfare
14000ms
62
E4
100ms
E4
100ms
E4
2000ms
fanfare
14000ms
63
F2
100ms
F2
100ms
F2
2000ms
white noise
14000ms
Unpredictable
Predictable
Unpredictable
Predictable
Predictable
Unpredictable
Detecting awareness from autonomic responses 27
64
G2
100ms
G2
100ms
G2
2000ms
fanfare
14000ms
65
E5
100ms
E5
100ms
E5
2000ms
fanfare
14000ms
Note. The piano notes are written with the octave number, for example D#2 refers to the note
‘D sharp’ played in the second octave. The duration of the test phase was approximately 20
minutes, starting 5 minutes after the training phase had completed. The delay after training
was included to reduce participants’ habituation to the aversive noise and the consequent
reduction in their SSR. The first trial – a predictable aversive trial – is included to
reacclimatise participants to the white noise after the delay; it is not included in the analyses.
The piano notes and the stimuli (fanfare or white noise) were played to different ears. In
blocks 1-4 the piano notes were played to the left ear and the stimuli to the right ear, in the
remaining blocks this was reversed.
Detecting awareness from autonomic responses 28
Visual Discrimination Task
The visual task was employed for the purpose of distraction in the unattended
condition. The task required participants to discriminate between four potential stimuli
occurring in random order at a rate of 1 Hz. The stimuli consisted of a 1 cm line oriented
either vertically or horizontally and positioned at a vertically central position either towards
the left or right side of the screen. In order to remove the potentially confounding effect of
movement artefacts on skin conductance measurements, responses were given using eye
movements. A sham eye-tracking camera was positioned on top of the screen, and
participants were told that this would record their eye-movements. They were instructed that
if a vertical line appeared on the right they were to briefly fixate the cross located at the top
of the screen, and if a horizontal line appeared on the left they were to briefly fixate the cross
located at the bottom of the screen, see Figure 1 below. For the two remaining alternatives, a
vertical line on the left or a horizontal line on the right, participants were instructed to
continue looking at the centre of the screen.
Right and Vertical
+
+
Left and Horizontal
Figure S1. Example of the screen layout used for the visual discrimination task employed in
the unattended condition.
Detecting awareness from autonomic responses 29
Experimental factors influencing rule acquisition
During development of the training sequence a variety of experimental factors were
found to influence the reliability with which participants acquired the rule. While these
factors were not systematically explored, the following provides a summary of those
identified to have a substantial influence together with the insights acquired from interviews
with participants of the pilot studies.
Richness of the stimuli. Piano notes were more effective than pure tones.
Participants’ comments indicated that this was based on the greater ease with which the pitch
of piano notes could be distinguished.
Delay between notes. A short delay (100 ms) between notes was found to be
preferable to longer delays (300+ ms). Participants’ reports suggest that the shorter delay
inclined them towards processing the pattern of three notes as a single predictive stimulus
rather than as separate notes.
Delay between the note pattern and subsequent stimulus. While there is a two
second delay between the piano notes and the subsequent stimulus during testing, a delay of
this length inhibited initial learning of the rule. Participants’ comments suggest that this effect
was due to a longer delay reducing awareness that the notes and subsequent stimulus formed
a single trial. This was addressed by having an initial delay of one second that incrementally
increased to two seconds by the end of the training phase.
Delay between trials. During testing the delay between trials was 14 seconds in order
to provide a suitable window for monitoring the sympathetic skin responses. However, delays
of this length during training reduced learning. An initially short delay of two and a half
seconds was ultimately found to be effective. Participants’ comments indicated that the
Detecting awareness from autonomic responses 30
longer delay reduced the likelihood that they would contrast one trial with the next and
consequently notice the predictive pattern.
Note repetition. Initial pilots included just two notes, with the second note either the
same or different from the first. However, learning was found to be more reliable if there
were three notes, with the final note being either the same or different. Participants indicated
that the repetition caused more anticipation of the final note thus making any difference more
salient.
Elimination of irrelevant variations. Variations in the stimuli that were not relevant
to the rule, particularly early in training, were found to substantially reduce learning. This
most obviously applied to note pitch. The predictive rule is independent of pitch i.e. three
notes the same irrespective of pitch predicts the fanfare, while two notes the same followed
by one of a different pitch predicts the noise burst. However, reliable learning of this rule was
only achieved when the training sequence was structured in stages. It was found to be
necessary for initial blocks of training trials to differ only in whether the last note was the
same or different, and not in pitch. This was repeated in subsequent blocks with each block
employing a different pitch. Only at the end of the training phase was a different pitch used
on each trial. When variations in pitch were introduced too early in training, participants were
found to generate incorrect inferences about those changes and fail to identify the intended
difference relating to the last of the three notes. A similar difficulty was identified with the
‘different’ third note. In early pilots the different note could be either higher or lower than the
previous two notes. However, this was again found to reduce learning of the rule, with
participants being inclined to hypothesise about consecutive trials showing the same or
different direction of pitch change. In the final sequence this is avoided by the ‘different’
third note always being lower than the previous two. It is always lower by greater than one
full octave to ensure the direction of the difference is clear.
Detecting awareness from autonomic responses 31
A fixed trial format. Initial pilots included a noise burst on aversive trials, where the
third note was different, and an equivalent period of silence on the non-aversive trials, where
the three notes were the same. However, it was found to be beneficial to include a ‘positive’
sound on non-aversive trials in place of the silence. Participants indicated that this made it
clearer that there was a fixed trial format, namely three notes followed by either a positive or
negative sound. Various positive sounds were piloted, including applause, a fanfare, and a
positive ‘ting’. The ‘fanfare’ was found to be most readily recognised as indicating a positive
outcome.
Delays between blocks of training trials. It was found to be beneficial to include
periodic pauses between blocks of training trials. These were inserted between blocks
employing the same note frequencies. Participants indicated that these pauses inclined them
to refocus attention on the task. If they were yet to identify any rules in the pattern the pauses
provided time for reflection and encouraged them to start afresh on the next block.
Interaural separation. It was apparent that despite the cues present in the stimuli
sequence, a small proportion of participants in the pilot studies failed to recognise that each
trial consisted of a predictive pattern (three notes) followed by a predicted event (noise burst
or fanfare). It was found that playing the two parts to separate ears increased the salience of
this conceptual separation.