Consciousness and Cognition 21 (2012) 408–412

Contents lists available at SciVerse ScienceDirect

Consciousness and Cognition

journal homepage: www.elsevier.com/locate/concog

Short Communication

Approach/avoidance in dreams
Susan Malcolm-Smith a,⇑, Sheri Koopowitz b, Eleni Pantelis b,c, Mark Solms b,c
a
ACSENT Laboratory, Dept. of Psychology, University of Cape Town, Rondebosch 7701, South Africa
b
Dept. of Psychology, University of Cape Town, South Africa
c
Dept. of Neurology, University of Cape Town, South Africa

a r t i c l e i n f o a b s t r a c t

Article history: The influential threat simulation theory (TST) asserts that dreaming yields adaptive advan-
Received 12 April 2011 tage by providing a virtual environment in which threat-avoidance may be safely
Available online 22 December 2011 rehearsed. We have previously found the incidence of biologically threatening dreams to
be around 20%, with successful threat avoidance occurring in approximately one-fifth of
Keywords: such dreams. TST asserts that threat avoidance is over-represented relative to other possi-
Threat simulation theory ble dream contents. To begin assessing this issue, we contrasted the incidence of ‘avoid-
Dream content
ance’ dreams with that of their opposite: ‘approach’ dreams. Because TST states that the
Approach
Avoidance
threat-avoidance function is only fully activated in ecologically valid (biologically threaten-
Mesocorticolimbic dopamine system ing) contexts, we also performed this contrast for populations living in both high- and low-
Reward seeking threat environments. We find that ‘approach’ dreams are significantly more prevalent
Amygdala across both contexts. We suggest these results are more consistent with the view that
Fear conditioning dreaming is generated by reward-seeking systems than by fear-conditioning systems,
Reward system although reward-seeking is clearly not the only factor determining the content of dreams.
Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction

The adaptive function of dreaming, if any exists, remains elusive. A recently influential hypothesis is the ‘threat-simula-
tion’ theory of Revonsuo (2000). According to this theory, dreaming provides a virtual environment in which biologically
threatening situations can be safely rehearsed: ‘‘the constant nocturnal rehearsing of threat perception and threat avoidance
skills increased the probability of successful threat-avoidance in real situations and thus led to increased reproductive fit-
ness’’ (Revonsuo, 2000, p. 898). This claim is consistent with the now well-established fact that the amygdala, a brain struc-
ture pivotally implicated in fear conditioning (Le Doux, 2002), is highly activated during REM sleep (Braun et al., 1997;
Maquet et al., 1996; Nofzinger, Mintun, Wiseman, Kupfer, & Moore, 1997). Revonsuo argues that the amygdala is activated
in dreams to evaluate hallucinated threats and assist in selecting the appropriate avoidance response (op cit, pp. 886–887,
894). However, as Revonsuo acknowledges, the amygdala is not the only instinctual-emotion circuit that is highly activated
during dreaming sleep; in fact the entire limbic system is activated (Braun et al., 1997; Maquet et al., 1996; Nofzinger et al.,
1997).
Despite the fact that research (reviewed by Revonsuo (2000)) indicates that dream content is often negative, and that
traumatic events impact on dream content, the focus on fear-conditioning in threat simulation theory (TST) has raised ques-
tions, particularly in light of low rates of real physical threat in dreams, and even lower rates of successful avoidance of such
threats (e.g. Malcolm-Smith & Solms, 2004; Malcolm-Smith, Solms, Turnbull, & Tredoux, 2008a; Zadra, Desjardins, &
Marcotte, 2006). It seems that around 20% of dreams (recent dreams of young adults, and recurrent dreams) feature realistic

⇑ Corresponding author. Fax: +27 21 650 4104.

E-mail addresses: Susan.Malcolm-Smith@uct.ac.za (S. Malcolm-Smith), Mark.Solms@uct.ac.za (M. Solms).

1053-8100/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.concog.2011.11.004
 S. Malcolm-Smith et al. / Consciousness and Cognition 21 (2012) 408–412 409

threats to the dreamer, while successful threat avoidance occurs in less than 5% of these reports (Malcolm-Smith & Solms,
2004; Malcolm-Smith et al., 2008a; Zadra et al., 2006). Likewise, although the incidence of threatening dreams is increased in
post-traumatic situations, evidence of successful threat-avoidance responses in such dreams is typically lacking (Punamaki,
1999; Valli, Revonsuo, Palkas, & Punamaki, 2006; Valli et al., 2005).
The authors of TST point out that even the low rates of dream threat evidenced above still suggest that the incidence of
threatening situations in dreams is far higher than that experienced in waking life (Valli & Revonsuo, 2009; Valli, Strand-
holm, Sillanmäki, & Revonsuo, 2008). These low percentages may also represent a disproportionately high incidence of
dreams with threatening content in relation to all possible instinctual-emotional contents. This raises the question of the
relative contributions to dream content of the various basic emotions. To begin to address this question, it is necessary to
compare the incidence of threat-avoidance behavior in dreams with that of at least one other instinctual-emotional behavior,
with an equivalent but different biological function. For this purpose we have selected exploratory-interest behavior, asso-
ciated with the mesocorticolimbic dopamine circuit variously described as the ‘reward’, ‘wanting’ or ‘SEEKING’ system
(Berridge, 1996; Panksepp, 1998; Rolls, 2000). Our rationale is that this system, like the amygdala, is highly activated during
REM sleep (Braun et al., 1997; Dahan et al., 2007; Léna et al., 2005; Maquet et al., 1996; Nofzinger et al., 1997), but it is asso-
ciated with behaviors that are for the most part antithetical to those associated with the amygdala. In essence, whereas the
amygdala (the FEAR system, in the nomenclature of Panksepp (1998)) is associated with ‘avoidance’ behavior, the SEEKING
system is associated with ‘approach’ behavior.
This ‘avoidance’ versus ‘approach’ dichotomy seems an appropriate place to begin to address the question of the relative
contributions to dream content of the different instinctual emotion systems, but it cannot by itself provide a definitive an-
swer. Several such systems exist in the mammalian brain, with more complex behavioral properties than mere avoidance or
approach, and there are overlapping spheres of influence between them (Panksepp, 1998). The relative contributions of all
these systems must ultimately be assessed in relation to each other. However, a rationale for comparing the incidence of
‘avoidance’ and ‘approach’ behaviors in dreams in this initial attempt to address the question, is the fact that Solms has
hypothesized that dreaming is generated by mesocorticolimbic dopamine activity, and in fact represents SEEKING activity
during sleep (Solms, 2000, 2011). Similar hypotheses have been advanced by others (Gottesmann, 2002; Panksepp, 1998;
Rotenberg, 1993). The comparison we have set for this initial study therefore serves the additional purpose of testing an
alternative prediction, arising from another recently influential dream theory.

2. Method

2.1. Sample

Purposive sampling was used to obtain dream reports from participants living in high versus low actual threat contexts,
namely undergraduate psychology students at the Universities of Cape Town and Wales (Bangor). The students were not
familiar with Revonsuo’s or Solms’s theories or with the aims of the study. Participation was voluntary and the research con-
formed to the ethical guidelines of both universities. Among the South African participants, 49.15% had experienced a real
life-threatening event in the past 4 years whereas the figure for the Welsh participants was 21.59% (Malcolm-Smith et al.,
2008a).

2.2. Procedure

The ‘Most Recent Dream’ method was used to obtain the dream reports. This method provides a representative sample of
dreams, indistinguishable in content (according to the measures of Hall and Van de Castle’s (1966)) from those obtained in
sleep laboratory settings (Domhoff, 1996). The questionnaire was distributed at the end of a lecture period. As is customary
with this method, reports of recurrent dreams and those containing 50 words or less were excluded from the analysis. This
yielded a sample of 208 South African and 116 Welsh dream reports, from which 105 reports were randomly selected from
each college population. This sample size is sufficient to replicate Hall and Van de Castle’s (1966) norms and thus is argued to
provide a reliable representation of dream content (Domhoff, 1996).
The reports were coded by three independent raters who were blind to the hypotheses of the study and the provenance of
the dream reports. Raters were asked to code the central events of dream reports in a forced-choice paradigm, as globally
involving either ‘avoidance’ or ‘approach’ behavior on the part of the reporting subject. ‘Avoidance’ behavior was defined
as: ‘the main activity of the subject of the dream is an attempt to avoid something through fleeing, freezing, hiding or the like’.
‘Approach’ behavior was defined as: ‘the main action of the subject of the dream is an attempt to approach something through
engagement, exploration, curiosity or the like’. Raters were instructed to code the dream in accordance with the dreamer’s
actual behaviors rather than their subjective feeling states in the dream (whether described or inferred). This was necessary
to take account of instances in which behaviors and feelings contradict each other. To enhance validity and reliability, coding
guidelines were provided, using characterizations of typical behaviors associated with the FEAR and SEEKING systems as
examples of ‘avoidance’ and ‘approach’ behaviors respectively (see Appendix A). These were derived from Panksepp
(1998) and Davies, Panksepp, and Normansell (2003). Before rating the 210 dream reports collected for this research, raters
were asked to rate a sample of twenty reports from another source obtained by the same method. The percentage of perfect
410 S. Malcolm-Smith et al. / Consciousness and Cognition 21 (2012) 408–412

agreement was calculated at 80%. This is a stringent measure of inter-rater reliability as it counts only absolute agreement
between all three raters on each individual report (Domhoff, 1996).

3. Results

Our participants were undergraduate psychology students, 18–25 years of age. The groups of South African and Welsh
students did not differ on age; t(208) = 1.63, p = .11. Both groups contained many more female than male students (as is com-
mon in psychology courses), and the gender distribution across the groups was equivalent; v2(1) = 0.19, p = .66. The groups
provided dream reports that were equivalent in length; M (SD) for South Africa = 129.42 (59.91); Wales = 110.31 (66.74),
t(208) = 1.68, p = .11.
Out of the total of 210 dream reports, significantly more were rated as instances of ‘approach’ (N = 153) than ‘avoidance’
(N = 57) behavior, v2(1) = 43.89, p < .0001. In the total sample, dream reports were thus nearly three times more likely to be
assigned to the ‘approach’ than the ‘avoidance’ category, Odd’s ratio = 2.7. This pattern was consistent for both the high and
low threat contexts, v2(1) = .22, p = .64 (i.e., the type of behavior predominant in dreams was not contingent on context; see
Table 1).

4. Discussion

These results indicate that manifest ‘approach’ behavior is more common than ‘avoidance’ behavior in dreams. This is
inconsistent with the view that the primary adaptive purpose of dreaming is fear-conditioning, but supports the hypothesis
that dreams represent reward-seeking activity. Taken together with our previously reported findings of a low incidence of
threat-avoidance in dreams – and specifically of successful threat-avoidance – even in actual threatening contexts (Malcolm-
Smith & Solms, 2004; Malcolm-Smith, Solms, Turnbull, & Tredoux, 2008b; Malcolm-Smith et al., 2008a) these results cast
doubt on the view that threat simulation is the primary function of dreaming.
However, several caveats apply. Although the ‘most recent dream’ method has repeatedly been shown to yield represen-
tative samples of dreams, indistinguishable from those obtained in sleep laboratory settings (see Domhoff, 1996), dream
content was coded according to Hall and Van de Castle (1966) criteria. It remains possible that dreams collected by the ‘most
recent dream’ method differ from those obtained in sleep laboratory settings when they are coded according to the ‘avoid-
ance’/’approach’ measure used here. This can only be determined by a comparative study using dream samples obtained by
both methods. Sleep laboratory and morning recall methods can also determine more precisely than the ‘most recent dream’
method the temporal relation between the reported dreams and preceding traumatic events (see Revonsuo & Valli, 2000;
Valli, Lenasdotter, MacGregor, & Revonsuo, 2007; Valli et al., 2008; also see Revonsuo and Valli (2008) and Valli and Revo-
nsuo (2009) for additional criticism of the ‘most recent dream’ method).
Of substantial theoretical importance is the further caveat that ‘avoidance’ as operationalized in this study is not synon-
ymous with ‘threat avoidance’ in the fullest sense implied by TST. Although the prototypic ‘avoidance’ responses associated
with amygdala activation are ‘flight’ and ‘freezing’ behaviors, an equally basic response to threat (also mediated by the
amygdala) is ‘fight’ behavior. In the simple dichotomy employed in this study, the latter response would have been coded
as an ‘approach’ behavior, in accordance with the instruction to rate the dream reports for manifest behaviors rather than
feeling states. This obscures the fact that attack behavior is often a defensive response to threat, which qualifies as an adap-
tive response of the kind predicted by TST. In other words, approach behaviors (and SEEKING activation) are not incompat-
ible with some forms of threat avoidance. These complexities can only be clarified by future studies which consider the
influence on dream content of a wider range of basic emotion command systems, and which consider not only manifest
dream behaviors but also subjective affects. Such studies could also refine the unit of analysis employed in this study, from
the central or focal event to the multiple behavioral sequences that typically occur in a dream.
Notwithstanding these caveats, however, the results of this study do provide support for the theory that dreaming repre-
sents SEEKING activity during sleep (Solms, 2000, 2011). These results, based on a content analysis method, may be added to
the converging lines of support for this theory based on a variety of other methods: dreaming is obliterated by lesions along the
path of the mesocorticolimbic dopamine system (Solms, 1997; Yu, 2001, 2007a); dreaming is enhanced by administration of
dopamine agonists (Hartmann, Russ, Oldfield, Falke, & Skoff, 1980); dreaming is suppressed by dopamine antagonists (Yu,
2007b); REM sleep is characterized by heightened mesocorticolimbic activity on PET (Braun et al., 1997; Maquet et al.,
1996; Nofzinger et al., 1997); REM sleep is characterized by prominent burst firing in the source cells of the mesocorticolimbic

Table 1
Incidence of ‘approach’ and ‘avoidance’ behavior in dreams.

Approach Avoidance
Total 72.9% (n = 153) 27.1% (n = 57)
S. Africa 74.3% (n = 78) 25.7% (n = 27)
Wales 71.4% (n = 75) 28.6% (n = 30)
 S. Malcolm-Smith et al. / Consciousness and Cognition 21 (2012) 408–412 411

dopamine system (Dahan et al., 2007); and REM sleep is characterized by maximal release of dopamine in the terminal nucleus
of this system (Léna et al., 2005).
It is important to recognize that the SEEKING theory and TST are not mutually exclusive. The authors of TST never claimed
that fear-conditioning was the only adaptive function of dreaming. Moreover it is common cause that foraging and other re-
ward-seeking exploratory behaviors are accompanied by constant vigilance, and therefore by co-activation of the amygdala
– which is in fact a major destination of mesocorticolimbic dopaminergic projections (Panksepp, 1998). We have also
acknowledged already that the SEEKING system energizes defensive attack behaviors. These facts are all consistent with
the observation that REM behavior disorder (when the behaviors are complex enough to be characterized) and the instinc-
tual behaviors of experimentally lesioned animals deprived of REM atonia are frequently characterized by freezing, fleeing
and fighting (Montplaisir, Gagnon, Postuma, & Vendette, 2011). Indeed, it has long been recognized that dopamine agonists
not only enhance dream frequency and intensity, but also increase nightmares (Hartmann et al., 1980; Nausieda, Weiner,
Kaplan, Weber & Klawans, 1982; Scharf, Moskowitz, Lupton, & Klawans, 1978). These L-DOPA induced dream changes typ-
ically herald psychosis; and most psychotic states are strongly characterized by fear and aggression. In short, excessive acti-
vation of mesocorticolimbic reward-seeking circuits typically culminates in negative affects. The role of anxiety in dreams
may be but a special instance of this general association.

Appendix A. Instructions to raters

Decide whether the main activity in the dream as a whole involves ‘avoidance’ or ‘approach’ behavior on the dreamer’s
part. The behavior must be rated either ‘avoidance’ or ‘approach’; there is no other alternative. If more than one event is de-
scribed, code the sequence of events as a whole.
‘Avoidance’ behavior is defined as: ‘the main activity of the subject of the dream is an attempt to avoid something through
fleeing, freezing, hiding or the like’. ‘Approach’ behavior is defined as: ‘the main action of the subject of the dream is an at-
tempt to approach something through engagement, exploration, curiosity or the like’.
Code the dream in accordance with the dreamer’s actual behavior rather than their feelings, even if these contradict each
other. For example, if the dreamer approaches an unknown place despite feeling scared, that is an instance of ‘approach’
behavior. Likewise, if the dreamer is curious about an unknown person but hides away from him/her, that is ‘avoidance’
behavior.
The following are prototypical examples of ‘avoidance’ and ‘approach’ behaviors, based on the scientific literature. These
examples are not meant to be exhaustive; they just describe good examples of the two types.
Typical examples of ‘approach’ behavior:

1. The dreamer engages with a thing/place/person/problem in an invigorated exploratory fashion, investigating or puzzling
over it or trying to make sense of it.
2. The dreamer acts in a persistently or intensely interested/curious/inquisitive fashion.
3. The dreamer is eagerly seeking new sensations or exciting experiences.
4. The dreamer is searching for something or pursuing a goal, even if s/he does not or cannot achieve it.
5. The dreamer acts as though s/he is looking forward to something and/or anticipating something.
6. Almost any little thing stimulates the dreamer’s interest.

Typical examples of ‘avoidance’ behavior:

1. The dreamer is acting in an apprehensive, tense, worried or generally nervous fashion.

2. The dreamer is attempting to escape and avoid something unpleasant.
3. The dreamer is frozen or rooted to the spot by something frightening.
4. The dreamer is stuck and cannot reach a decision about something.
5. The dreamer misses an opportunity due to worry or anxiety.
6. The dreamer acts as though s/he dreads something bad.

References

Berridge, K. C. (1996). Food reward: Brain substrates of wanting and liking. Neuroscience and Biobehavioural Reviews, 20, 1–25.
Braun, A. R., Balkin, T., Wesensten, N. J., Carson, R. E., Varga, M., Baldwin, P., et al (1997). Regional cerebral blood flow throughout the sleep-wake cycle. An
H215O PET study. Brain, 120, 1173–1197.
Dahan, L., Astier, B., Vautrelle, N., Urbain, N., Kocsis, B., & Chouvet, G. (2007). Prominent burst firing of Dopaminergic neurons in the ventral tegmental area
during paradoxical sleep. Neuropsychopharmacology, 32, 1232–1241.
Davies, K. L., Panksepp, J., & Normansell, L. (2003). The affective neuroscience personality scales: Normative data and implications. Neuropsychoanalysis, 5,
57–69.
Domhoff, G. W. (1996). Finding meaning in dreams: A quantitative approach. New York: Plenum Press.
Gottesmann, C. (2002). The neurochemistry of waking and sleeping mental activity: The disinhibition-dopamine hypothesis. Psychiatry and Clinical
Neurosciences, 56, 345–354.
Hall, C., & Van de Castle, R. (1966). The content analysis of dreams. New York: Appleton-Century-Crofts.
412 S. Malcolm-Smith et al. / Consciousness and Cognition 21 (2012) 408–412

Hartmann, E., Russ, D., Oldfield, M., Falke, R., & Skoff, B. (1980). Dream content: Effects of L-Dopa. Sleep Research, 9, 153.
Le Doux (2002). Synaptic self: How our brains become who we are. New York: Viking.
Léna, I., Parrot, S., Deschaux, O., Muffat-Joly, S., Sauyinet, V., Renaud, B., et al (2005). Variations in extracellular levels of dopamine, noradrenaline, glutamate,
and aspartate across the sleep-wake cycle in the medial prefrontal cortex and nucleus accumbens of freely moving rats. Journal of Neuroscience Research,
81, 891–899.
Malcolm-Smith, S., & Solms, M. (2004). Incidence of threat in dreams: A response to Revonsuo’s threat simulation theory. Dreaming, 14, 220–229.
Malcolm-Smith, S., Solms, M., Turnbull, O., & Tredoux, C. (2008a). Threat in dreams: An adaptation? Consciousness and Cognition, 17, 1281–1291.
Malcolm-Smith, S., Solms, M., Turnbull, O., & Tredoux, C. (2008b). Shooting the messenger won’t change the news. Consciousness and Cognition, 17,
1297–1301.
Maquet, P., Péters, J., Aerts, J., Delfiore, G., Dequeldre, C., Luxen, A., et al (1996). Functional neuroanatomy of human rapid-eye-movement sleep and
dreaming. Nature, 383, 163–166.
Montplaisir, J., Gagnon, J-F., Postuma, R., & Vendette, M. (2011). REM sleep parasomnias. In P. Montagna & S. Chokroverty (Eds.). Handbook of clinical
neurology, sleep disorders – Part 1 (Vol. 99, 3rd series, pp. 869–882). New York: Elsevier.
Nausieda, P. A., Weiner, W. J., Kaplan, L. R., Weber, S., & Klawans, H. L. (1982). Sleep disruption in the course of chronic levodopa therapy: An early feature of
the levodopa psychosis. Clinical Neuropharmacology, 5, 183–194.
Nofzinger, E. A., Mintun, M. A., Wiseman, M., Kupfer, D. J., & Moore, R. Y. (1997). Forebrain activation in REM sleep: An FDG PET study. Brain Research, 770,
192–201.
Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. New York: Oxford University Press.
Punamaki, R-L. (1999). The relationship of dream content and changes in daytime mood in traumatized vs. non-traumatized children. Dreaming, 9, 213–234.
Revonsuo, A. (2000). The reinterpretation of dreams: An evolutionary hypothesis of the function of dreaming. Behavioral and Brain Sciences, 23, 877–901.
Revonsuo, A., & Valli, K. (2000). Dreaming and consciousness: Testing the threat simulation theory of the function of dreaming. Psyche, 6. <http://
psyche.cs.monash.edu.au/v6/psyche-6-08-revonsuo.html>. Retrieved 22.12.03.
Revonsuo, A., & Valli, K. (2008). How to test the threat-simulation theory. Consciousness and Cognition, 17, 1292–1301.
Rolls, E. T. (2000). The orbitofrontal cortex and reward. Cerebral Cortex, 10, 284–294.
Rotenberg, V. (1993). REM sleep and dreams as mechanism of search activity recovery. In A. Moffitt, M. Kramer, & R. Hoffmann (Eds.), Function of dreaming
(pp. 261–292). New York: New York State University of New York Press.
Scharf, B., Moskowitz, C., Lupton, M., & Klawans, H. (1978). Dream phenomena induced by chronic Levodopa therapy. Journal of Neural Transmission, 43,
143–151.
Solms, M. (1997). The neuropsychology of dreams: A clinico-anatomical study. New Jersey: Lawrence Erlbaum.
Solms, M. (2000). Dreaming and REM sleep are controlled by different brain mechanisms. Behavioral and Brain Sciences, 23, 843–850.
Solms, M. (2011). Neurobiology and the neurological basis of dreaming. In P. Montagna & S. Chokroverty (Eds.). Handbook of clinical neurology, sleep
disorders – Part 1 (Vol. 98, 3rd series, pp. 519–544). New York: Elsevier.
Valli, K., Lenasdotter, S., MacGregor, O., & Revonsuo, A. (2007). A test of the threat simulation theory: Replication of results in an independent sample. Sleep
and Hypnosis, 9, 30–46.
Valli, K., & Revonsuo, A. (2009). The threat simulation theory in light of recent empirical evidence: A review. American Journal of Psychology, 122, 17–38.
Valli, K., Revonsuo, A., Palkas, O., Ismail, K., Ali, K., & Punamaki, R. (2005). The threat simulation theory of the evolutionary function of dreaming: Evidence
from dreams of traumatized children. Consciousness and Cognition, 14, 188–218.
Valli, K., Revonsuo, A., Palkas, O., & Punamaki, R. (2006). The effect of trauma on dream content – A field study of Palestinian children. Dreaming, 16, 63–87.
Valli, K., Strandholm, T., Sillanmäki, L., & Revonsuo, A. (2008). Dreams are more negative than real life: Implications for the function of dreaming. Cognition &
Emotion, 22, 833–861.
Yu, C. K. (2001). Neuroanatomical correlates of dreaming II. The ventromesial frontal region controversy (dream instigation). Neuropsychoanalysis, 3,
193–201.
Yu, C. K. (2007a). Cessation of dreaming and ventromesial frontal-region infarcts. Neuropsychoanalysis, 9, 83–90.
Yu, C. K. (2007b). Brain mechanisms of dreaming. Unpublished PhD thesis, University of Cape Town.
Zadra, A., Desjardins, S., & Marcotte, E. (2006). Evolutionary function of dreams: A test of the threat simulation theory in recurrent dreams. Consciousness and
Cognition, 15, 450–463.