Awareness is awareness is awareness? Decomposing different aspects of awareness and their role in operant learning of pain sensitivity
Highlights
► Different aspects of awareness were distinguished. ► Implicit operant learning modulated pain sensitivity. ► Contingencies and reinforcing stimuli do not have necessarily to be discriminated. ► Contingencies had neither to be discriminated in objective nor in a subjective test. ► Explicit cognitive processes impaired implicit learning of enhanced habituation.
Introduction
Implicit learning is described as learning without awareness. Although the existence of implicit learning has been debated at large (Lovibond and Shanks, 2002, Shanks, 2005), its existence is now widely accepted (Cleeremans et al., 1998, Kirsch et al., 2004). However, there is no generally accepted way of determining which learning processes are implicit. One reason for this lack of agreement, and the controversy in general, is the view of awareness as a consistent and universal concept. Awareness consists of different aspects, which often receive little attention. For example, awareness can refer to sensory awareness of a stimulus or to contingency awareness, i.e. being aware of a conditional connection. One approach to differentiate aspects of awareness is to decompose knowledge into ‘structural knowledge’ and ‘judgment knowledge’. Structural knowledge refers to knowledge about the structure of items, their relationships, etc., while ‘judgment knowledge’ refers to the content directly expressed in a judgment, in which a structure is present. Therefore, judgment knowledge is based on structural knowledge, and both judgment and structural knowledge can be conscious or unconscious (Dienes, 2008, Dienes and Scott, 2005). Results in an implicit learning task (artificial grammar learning task) suggest that conscious and unconscious structural and judgment knowledge are qualitatively different and have different significance for the learning process (Dienes & Scott, 2005). In general, it is conceivable that learning remains unaffected by some aspects of awareness while others may be essential to learning. However, explicit processes like cognition, even if not directly related to the learning process, can have beneficial as well as detrimental effects on implicit learning (c.f. Kirsch et al., 2004, Sun et al., 2005). With the present study we aimed to decompose different aspects of awareness and to test their role in an operant learning process as well as to assess possible interactions of this learning process with cognitive processes.
It is widely accepted that implicit and explicit learning result in different memory systems, namely non-declarative and declarative memory, which depend on different brain structures (Eichenbaum and Cohen, 2004, Squire, 2004). Implicit learning is regarded as learning without awareness and explicit learning as learning with awareness (Cleeremans et al., 1998). According to this description, awareness during learning indicates the underlying mechanism and memory system. However, it remains unclear what the concept of awareness implies and if testing different aspects of awareness leads to the same conclusion.
An example in which awareness was specified, leading consequently to clear conclusions about underlying learning mechanisms and memory functions, was a series of studies on classical eyeblink conditioning. In these studies, non-declarative and declarative memory functions were dissociated by analyzing the role of contingency awareness in delay and trace conditioning (Clark, Manns, & Squire, 2002). In delay conditioning the conditioned stimulus (CS) begins before and remains during the presentation of the unconditioned stimulus (US), both stimuli terminate simultaneously. Successful delay conditioning is independent of contingency awareness, i.e. recognizing the CS–US relation (Clark and Squire, 1999, Smith et al., 2005). Moreover, contingency awareness was demonstrated to be epiphenomenal in delay conditioning (Clark and Squire, 1999, Smith et al., 2005). In trace conditioning, the CS is presented and terminated prior to the US; a silent (trace) interval lies between the two stimuli. In contrast to delay conditioning, successful trace conditioning requires awareness of the US–CS contingency (Clark et al., 2002, Manns et al., 2000a, Manns et al., 2000b). Similar results were obtained in fear and aversive classical conditioning (Carter et al., 2006, Knight et al., 2004, Weike et al., 2007).
These studies demonstrated that awareness can indeed be used as an indicator for an underlying learning mechanism, as long as specific aspects of awareness are precisely defined and their significance for the learning process is tested. In associative learning—comprising classical and operant conditioning—awareness of the procedure being a learning task, awareness of the US or the reinforcing stimuli and contingency awareness have to be distinguished since they are qualitatively different.
The controversy concerning the existence of implicit learning has been further complicated by the methodological problem of assessing awareness. In general, subjective and objective methods have been used. Subjective methods—e.g. verbal reports or confidence ratings—demand the report of an internal state and thus introspection (Cleeremans et al., 1998). Due to this introspection, subjective methods have been considered as indexing “awareness of awareness” (Persaud et al., 2007, Wiens and Ohman, 2002). In contrast, objective methods infer awareness from a test behavior, e.g. recognition, prediction, or post-decision wagering (Cleeremans et al., 1998, Manns et al., 2000b, Persaud et al., 2007). Objective methods need no introspection and thus can be regarded as tests for (behavioral) first-order discrimination. Following this conception, subjective methods can be regarded as tests for (verbal) second-order discrimination (Dienes, 2008). Thus subjective and objective methods target different levels of processing. Further, this distinction reveals a problem with objective methods; first-order discrimination does not necessarily equal subjective awareness since behavioral discrimination is possible without being able to report the discriminative performance (Möltner and Hölzl, 2002, Reber and Squire, 1994, Wiens and Ohman, 2002). Therefore, these methods have been criticized as being insensitive to phenomenal experience (Kunimoto et al., 2001, Seth, 2008). Subjective methods, on the other hand, have been criticized as lacking sensitivity and hence prone to response biases (Cowey, 2004, Shanks, 2005). Consequently, it appears to be preferable to combine objective and subjective methods when assessing awareness.
Most studies on awareness in operant learning focused on contingency awareness—i.e. awareness of the conditional connection between a specific behavior and subsequent reinforcement—and applied subjective methods. However, results are inconsistent; contingency awareness has been demonstrated to be essential (e.g. DeNike and Leibovitz, 1969, Page and Lumia, 1968, Rescorla, 1991, Wearden and Shimp, 1985) as well as unnecessary for operant learning (e.g. Dienes, 2008, Hefferline and Keenan, 1963, Lieberman et al., 1998, Rosenfeld and Baer, 1970, Svartdal, 1992). Different definitions and criteria for awareness contributed to these inconsistent results (Cleeremans et al., 1998, Lovibond and Shanks, 2002). Therefore, the aim of the present study was to decompose different aspects of awareness and test their role in an operant learning task. The various aspects of awareness distinguished were contingency awareness, awareness of the procedure being a learning task, and awareness of the reinforcing stimuli. Since the role of contingency awareness is generally emphasized in associative learning, this aspect was assessed with a subjective and an objective method. It was further tested whether cognition on one’s own behavior affected the learning process, providing a possible explanation for the inconsistent results on awareness in operant learning since it is known that explicit processes can affect (implicit) learning (Sun et al., 2005).
The applied operant learning task was to modulate pain sensitivity. Such a modulation of pain sensitivity by operant learning, leading to hypersensitivity, has been demonstrated previously (Becker et al., 2008, Hölzl et al., 2005). Hypersensitivity in terms of sensitization is a clinical marker of chronic pain (Kleinböhl et al., 1999, Woolf and Salter, 2000). Whether hypersensitivity can be learned implicitly is of special interest in the context of chronic pain. For example, since implicit learning can lead to cumulative changes in perceptual systems (Clark & Squire, 1998), it provides an explanation for the development of hypersensitivity and chronic pain. Thus, it should be determined whether implicit operant learning can modulate pain sensitivity.
We hypothesized that operant learning of changes in pain sensitivity was independent of first- and second-order discrimination of the contingencies and thus of contingency awareness. Further, we hypothesized that awareness of the procedure being a learning task and awareness of the reinforcing stimuli were not necessary for successful operant conditioning. Explicit processes like cognition about one’s own behavior and its effect on the learning process were hypothesized not to affect implicit learning.
Section snippets
Participants
Thirty-three healthy volunteers (29 female; aged 23–64 years; M = 47.4, SD = 9.5) participated in the study. Participants were excluded if they reported pain episodes exceeding 1 day per month and neurological or psychiatric disorders. None of the healthy participants had to be excluded because of thresholds for warm and phasic pain outside a 95% norm range (norm data according to Rolke et al. (2006)). Each participant attended two experimental sessions, resulting in 66 sessions for 33 participants.
Operant modulation of pain sensitivity
The operant learning task was effective in modulating pain sensitivity according to the learning conditions. The effect of the operant contingencies was observable in (a) the local operant behavior within each trial, i.e. the temperature regulations by the participant and (b) globally in the long-term temperature change from baseline over the course of the conditioning session, resembling a learning curve composed of the accumulated local operant behavior.
The sum of up- and down-regulations of
Discussion
The present study emphasized the importance of decomposing different aspects of awareness as well as different levels of processing in order to determine whether a learning process is implicit. Taken together, these separate aspects can then converge at a conclusion concerning underlying mechanisms and memory systems of a learning process.
By decomposing awareness into awareness of the procedure being a learning task, awareness of the reinforcing stimuli, and awareness of the contingencies the
Acknowledgments
We want to thank Petra Schweinhardt for her helpful comments on a previous version of the manuscript. The research was supported by grants from the German Research Foundation (Clinical Research Unit 107: ‘Neuronal plasticity and learning in chronic pain’, Project Ho 904/11–1-4), by funds from the Otto Selz Institute for Applied Psychology, University of Mannheim, and by an unconditional grant from the ‘Berufsgenossenschaft Nahrungsmittel und Gaststätten’ (BGN; German statutory health and safety
References (60)
- et al.
Operant conditioning of enhanced pain sensitivity by heat-pain titration
Pain
(2008) - et al.
Quantitative sensory examination of epidural anaesthesia and analgesia in man: Effects of pre- and post-traumatic morphine on hyperalgesia
Pain
(1994) - et al.
Contingency awareness in human aversive conditioning involves the middle frontal gyrus
NeuroImage
(2006) - et al.
Classical conditioning, awareness, and brain systems
Trends in Cognitive Sciences
(2002) - et al.
Implicit learning: News from the front
Trends in Cognitive Sciences
(1998) - et al.
Instrumental responding remains under the control of the consequent outcome after extended training
Behavioural Processes
(2002) - et al.
The effect of pre- versus postinjury infiltration with lidocaine on thermal and mechanical hyperalgesia after heat injury to the skin
Pain
(1993) - et al.
A multiple random staircase method of psychophysical pain assessment
Pain
(1988) - et al.
Implicit operant learning of pain sensitization
Pain
(2005) - et al.
Psychophysical measures of sensitization to tonic heat discriminate chronic pain patients
Pain
(1999)