An auditory perceptual learning paradigm was used to investigate whether implicit memories are formed during general anesthesia.
Eighty-seven patients who had an American Society of Anesthesiologists physical status of I-III and were scheduled to undergo an elective surgery with general anesthesia were randomly assigned to one of two groups. One group received auditory stimulation during surgery, whereas the other did not. The auditory stimulation consisted of pure tones presented via headphones. The Bispectral Index level was maintained between 40 and 50 during surgery. To assess learning, patients performed an auditory frequency discrimination task after surgery, and comparisons were made between the groups. General anesthesia was induced with thiopental and maintained with a mixture of fentanyl and sevoflurane.
There was no difference in the amount of learning between the two groups (mean +/- SD improvement: stimulated patients 9.2 +/- 11.3 Hz, controls 9.4 +/- 14.1 Hz). There was also no difference in initial thresholds (mean +/- SD initial thresholds: stimulated patients 31.1 +/- 33.4 Hz, controls 28.4 +/- 34.2 Hz). These results suggest that perceptual learning was not induced during anesthesia. No correlation between the bispectral index and the initial level of performance was found (Pearson r = -0.09, P = 0.59).
Perceptual learning was not induced by repetitive auditory stimulation during anesthesia. This result may indicate that perceptual learning requires top-down processing, which is suppressed by the anesthetic.
GENERAL anesthesia (GA) suppresses conscious perception. However, the brain is not completely shut down during anesthesia. For example, sensory stimulation still evokes a strong first wave of cortical activation.1Even after a century of research, it is still largely unknown to what extent stimulation can induce learning and memory. For example, words presented during GA influence postsurgery reading performance2and word-stem completion.3However, it remains unknown whether the underlying learning is implicit or related to explicit learning during short periods of “lighter” anesthesia.2,4–10
A reliable paradigm to investigate implicit learning is perceptual learning. In perceptual learning, repetitive stimulation with the same stimuli, e.g. , two tones, which are hard to discriminate, yields improvement of performance, i.e. , it becomes easier to discriminate the two tones. With this training, performance improves slowly in hundreds of stimuli presentations.11Hence, learning is not caused by very few presentations of stimuli, as is the case with, for example, word stem completion. Crucial for the current study is that perceptual learning occurs in awake, healthy participants also when the stimuli are unattended or subliminal.12–17For example, Amitay et al. 16presented tones of different frequencies to participants who performed tasks unrelated to the tones (e.g. , reading a book). Even though the tones were not attended, the passive stimulation improved the frequency discrimination of the tones. Therefore, the mere presentation of tones induced learning in these passive participants. Analogous results were also found in the visual domain, even with subliminal stimuli.14,15,17
Hence, if repetitive stimulation with unattended and subliminal stimuli is sufficient to provoke implicit learning in healthy, awake participants, it may also be possible to induce perceptual learning during GA.
To test this hypothesis, a group of anesthetized patients received auditory stimulation during surgery, whereas another group did not. Both groups performed an auditory frequency discrimination task after the surgery, and the performance of the groups was compared. If perceptual learning can be induced during GA, the group that received stimulation during surgery would perform the task better, as compared with the control group. If GA prevents perceptual learning, there should be no difference in performance between the groups. We found the latter to be true.
Materials and Methods
The study was approved by the local institutional ethics committee (University of Lausanne, Lausanne, Vaud, Switzerland). Patients who had an American Society of Anesthesiologists physical status of I–III and were aged 19–61 yr were recruited and provided informed written consent. These patients underwent visceral or orthopedic surgery during GA lasting for more than 60 min. In the preoperative phase, potential participants were tested to make sure they were eligible for the study. Exclusion criteria were hearing deficits, neurologic pathology, chronic alcoholism, a history of drug abuse, and treatment with benzodiazepines, neuroleptics, antidepressants, or antiepileptics, as well as not comprehending the task with very easy stimuli or having a hearing threshold above 30 dB for detecting a pure tone of 1 kHz. To investigate whether musical expertise had any influence on learning, patients had to indicate for how long they had played a musical instrument. Finally, 87 patients fulfilled the above criteria and were randomly assigned to one of two groups based on a premade list of patients. One of the groups received auditory stimulation during surgery, whereas the other, the control group, did not. A list was created by concatenating smaller lists containing 10 patients. In each of these lists, patients were randomly assigned to receive stimulation (n = 5) or to be controls (n = 5). The 1st list contained patients 1–10, the 2nd list contained patients 11–20, and so on.
In a pilot study, eight paid students from the Ecole Polytechnique Fédérale de Lausanne (Lausanne, Switzerland) were recruited (and performed the learning experiment without GA).
A schematic of the general procedure is illustrated in figure 1. Patients were told they would be placed randomly in one of two groups, and that one of these groups would receive auditory stimulation during the surgery. On the day of surgery, presentation of stimuli was initiated (for the stimulated group only) as soon as the Bispectral Index (BIS) value was between 40 and 50. After recovery, usually 24–48 h after surgery, all patients performed 10 blocks of the frequency discrimination task.
After arrival in the operating room, standard monitoring was applied. The forehead skin was cleaned with ether to improve skin conductance, and the BIS XP-Sensor® electrodes (Aspect Medical Systems, Newton, MA) were applied as recommended by the manufacturer. The electroencephalogram was recorded continuously using the Aspect A-2000 BIS® monitor (version XP).
After preoxygenation with 100% oxygen, anesthesia was induced with 2–3 μg/kg fentanyl, 3–6 mg/kg thiopental, and 0.1 mg/kg vecuronium. All patients were intubated by the oral route. Anesthesia was maintained with fentanyl, a mixture of air and oxygen, sevoflurane, and vecuronium if necessary. The in-charge physician was requested to maintain a BIS value between 40 and 50.
Intrasurgical Stimulation (Stimulated Group)
During each trial, two pure tones of 100 ms with 20 ms cosine rise/fall were presented in two intervals 500 ms apart. Tones were presented at 80 dB. One of the tones had a frequency of 1 kHz, whereas the other had a frequency of 1 kHz +Δf (with Δf being 1, 4, 16, or 64 Hz). The reference tone of 1 kHz was presented randomly in the first or the second interval. In each trial, one of the four levels of Δf was chosen randomly. All levels of Δf were presented equally many times. In total, patients were presented with 1,000 pairs of tones during surgery. Each such trial was separated by 1,500 ms. The entire stimulation lasted approximately 30 min. In a pilot study, perceptual learning was successfully induced in healthy students when using this paradigm (fig. 2A). Crucial for this study, passive stimulation with pure tones induced learning in a frequency discrimination task for awake participants.16
Postsurgical Discrimination Task (Stimulated and Control Groups)
In the postsurgical frequency discrimination task, participants fixated a white cross in the middle of a screen while auditory stimuli were presented via headphones. Stimulation was identical to the intrasurgical stimulation, except that Δf was varied using an adaptive staircase method.18The task of the participants was to determine whether the tone with the higher frequency (1 kHz +Δf) was presented in the first (fig. 1A) or the second interval (fig. 1A). Participants responded by pressing one of two buttons. Visual feedback (the words Correct and Wrong in green and red letters, respectively) was provided after each response. Participants performed 10 blocks of 100 trials each. There was a mandatory pause of 10 min after the fifth block. By adaptively varying the difference in frequency, Δf, a discrimination threshold could be determined for each block. Learning occurs when the discrimination threshold becomes smaller (i.e. , participants can discriminate smaller Δf). Thresholds above 200 Hz were excluded to avoid outliers in the averaged data. With this paradigm, performance improved quickly.19
For the auditory stimulation before, during, and after surgery, tones were presented via closed headphones (Sennheiser HD 280 Pro; Sennheiser electronic GmbH & Co. KG, Wennebostel, Germany) connected to a laptop computer (Dell Latitude D410; Dell Inc., Austin, TX).
Data Analysis of Perceptual Learning
For each block, a discrimination threshold was determined by fitting a cumulative gaussian function to the obtained data. A criterion of 75% correct was used to determine the threshold. Therefore, for each patient, 10 discrimination thresholds were calculated.
To investigate differences in the amount of learning induced by performing the frequency discrimination task, a 2-way mixed-factors analysis of variance was used. There was a between-patients factor Group (stimulated, control) and a within-patients factor Time (T1, T2). T1 is the average discrimination threshold calculated over blocks 1–5, whereas T2 is the average discrimination threshold calculated over blocks 6–10 (fig. 1D).
To investigate differences in learning induced by auditory stimulation during surgery, the initial performances of the stimulated patients and of the control patients were compared. The initial performance was determined by calculating the average of the first three thresholds for each patient (fig. 1D). Statistical comparisons were performed with two-tailed permutation tests. Demographic data were compared using two-tailed permutation tests.
To investigate whether the change in performance was influenced by the level of anesthesia, the correlation between the initial performance levels of patients and their BIS values was determined using the Pearson product–moment coefficient.
Permutation tests require no assumptions about the underlying probability distributions.20,21The permutation test assumes that the null hypothesis is true (i.e. , there is no difference between the groups). One sample from the null distribution can be created by randomly shuffling patients between the two groups. For each sample, the statistic of interest is calculated, e.g. , the difference in age between groups. To determine whether the original sample is significantly different from the permutation distribution, the difference in age for the original sample is compared with each of the samples in the permutation distribution. Here, each permutation distribution consisted of 100,000 permuted samples. Data analysis was conducted using R, an open-source software package.22
One hundred four patients were initially included in the study and underwent a surgical procedure with GA. Thirteen patients declined to perform the postsurgical testing. Two more patients were excluded because of failure of performing the frequency discrimination task, and another two patients were excluded because of software failures. Therefore, 87 patients completed the study. Eighty-five percent of the BIS readings fell into the intended range (BIS 40–50), suggesting that patients were under deep anesthesia during stimulation. Patient characteristics are shown in table 1.
Learning curves and mean thresholds are shown in figures 2A and B. Participants in the pilot study improved performance with training (P < 0.01). To investigate differences in the amount of learning for the patient groups, a two-way mixed-factors analysis of variance with factors Group and Time was calculated (fig. 2B). There was a main effect of Time (F1,85= 45.4, P < 0.0001) but no main effect of Group (F1,85= 0.01, P = 0.91). The interaction Group × Time was not significant (F1,85= 0.01, P = 0.94). Hence, both groups of patients improved performance significantly in the frequency discrimination task. Because the performance of both groups was comparable (fig. 2B), the amount of learning was not influenced by the auditory stimulation during surgery.
This argument is also supported by the fact that the initial performance levels of both groups are roughly the same (fig. 2C). If the stimulation during surgery had induced learning, the initial performance of the stimulated group should be better compared with the control group. However, there was no difference in initial thresholds (mean initial threshold stimulated = 31.1 Hz, controls = 28.4 Hz; P = 0.72; fig. 2C).
It has been reported that implicit learning is correlated with the level of anesthesia. Patients with lighter levels of anesthesia improved performance more strongly.2,9To investigate the relation between the level of anesthesia and perceptual learning, the correlation between the initial performance and the BIS values was calculated for the stimulated patients only. Three patients were excluded because of initial thresholds being larger than 100 Hz (fig. 2D). There was no significant correlation, suggesting the level of anesthesia had no influence on the amount of learning (fig. 2D, solid line; Pearson r =−0.09, P = 0.59; if the patients with initial thresholds larger than 100 Hz were included in the analysis, correlations were still not significant; fig. 2D, dashed line; Pearson r =−0.17, P = 0.25).
In auditory perceptual learning, improvement of performance is not always specific to the trained stimuli.23Therefore, patients with musical experience might already be experts in performing the task. Indeed, the initial performance level of patients with at least 1 yr of practice playing a musical instrument was significantly better compared with patients without musical experience (fig. 3A; mean initial threshold musical experience = 15.1 Hz, no musical experience = 43.3 Hz; two-tailed permutation test: P < 0.00001). To conclude that the previous results were not due to ceiling effects of patients with musical experience, an additional analysis was performed only for patients reporting having no musical experience (patients who reported never having practiced playing an instrument). Figure 3Bshows the thresholds for the first and second parts of the experiment (i.e. , T1 and T2) for patients with no musical experience. A two-way analysis of variance with factors Group and Time was calculated for patients with no musical experience. There was a main effect of Time (F1,43= 43.5, P < 0.0001) but no main effect of Group (F1,43= 0.01, P = 0.94) or any significant interaction Group × Time (F1,43= 0.29, P = 0.59). Therefore, both groups improved performance when performing the frequency discrimination task, and there was no difference in the amount of improvement between groups. To compare initial thresholds, a two-tailed permutation test was conducted on the initial performance. Again, there was no difference in initial thresholds (fig. 3C; mean initial threshold stimulated = 44.6 Hz, controls = 41.4 Hz; P = 0.80). In addition, there was no significant correlation between the initial performance and the level of anesthesia (fig. 3D, solid line; Pearson r =−0.05, P = 0.83; including the three patients with initial thresholds above 100 Hz did not change the results; fig. 3D, dashed line; Pearson r =−0.21, P = 0.30).
Perceptual learning is a form of long-term implicit learning lasting for years.24Perceptual learning can be induced by subliminal stimuli in healthy, awake participants,14,15,17,e.g. , when paired with another task14,15or even with a reward signal only.17Moreover, perceptual learning can be induced without attending to stimuli, i.e. , passive stimulation.12,13,16For example, repetitive stimulation of pure tones with the same stimulus paradigm successfully induced perceptual learning in awake, healthy participants performing a different task.16Hence, we hypothesized that repetitive stimulation during GA may also induce perceptual learning. However, this was not the case. Overall, both patient groups improved performance equally on the postsurgical frequency discrimination task (figs. 2A and B). Also, the initial performance level was comparable between the groups (figs. 2B and C). Therefore, repetitive, auditory stimulation during anesthesia did not induce perceptual learning.
Awareness of Stimuli
Learning might be related to the level of anesthesia and awareness of stimuli.2–4,6,8,9Only a BIS level above 40 enabled formation of implicit memories in a word-stem completion task3and similar results were found in a speed-reading task.2In contrast, no significant relation between the level of anesthesia (BIS) and the initial performance levels was found in the current study (figs. 2D and 3D).
It has been suggested that during GA, patients are unaware of the stimuli and are therefore unable to form memories.25However, perceptual learning can be induced even when awake participants are unaware of the stimulation.14,15In addition, cortical plasticity can be induced by repetitive stimulation during anesthesia in rats.26Therefore, being unaware of the stimulation alone cannot explain why perceptual learning was not induced in the current study. Hence, GA might suppress perceptual learning itself or its consolidation.27
In general, learning occurs already as training proceeds, i.e. , it does not only show up after consolidation. This is evident also in the current study (figs. 2A and B; see also Amitay et al. 16and Hawkey et al. 19). Still, there is the possibility that successful learning, induced during GA, was erased by postsurgical conditions such as stress and sleep deprivation. However, although stress might have a profound effect on learning,28implicit learning is not sensitive to stress.29–31Furthermore, although sleep might be crucial for visual perceptual learning,32,33this is not the case for auditory learning.19,34,35Hence, it seems that perceptual learning was obliterated because of GA itself and not postsurgical conditions.27
Specificity and Nonspecificity of Perceptual Learning
Various studies show that GA influences the neural response in a nonspecific way.36–38Therefore, stimulation during GA might trigger different neural populations compared with the postsurgery testing. Perceptual learning is often very specific to the basic characteristics of the trained stimuli.39,40Hence, if different neural populations are triggered during GA compared with the awake state, plastic changes induced by the stimulation may not be beneficial in the awake state, i.e. , when performing a postsurgical task. However, the stimuli used in the current study are known to transfer across broad frequency bands, i.e. , they are less specific than normal visual stimuli (for this reason they were chosen in this study). Moreover, cortical plasticity was induced in rats with GA.26
It has been suggested that consciousness relies on bidirectional integration of sensory information1,41–43and that the cause of anesthetic-induced unconsciousness is an alteration of the information flow between higher-order areas and areas processing basic sensory information.1,44–47For example, GA reduces the information flow from frontal to posterior areas but preserves it from posterior to frontal areas.45–47Therefore, feed forward processing of stimuli seems to be intact during GA, whereas top-down processing is disrupted. Top-down influences have profound effects on perceptual learning and cortical plasticity,48–50something also acknowledged by computational models.51–53Importantly, also perceptual learning without awareness of stimuli seems to rely on top-down influences, e.g. , in the form of reward processing and feedback16,54(see introduction). Recently, it was shown that monkeys' improvement on a visual task could be related to changes in the neural response in early visual areas.55Interestingly, when the monkeys were anesthetized, there was no longer any difference in the neural response, thus suggesting that crucial top-down influences had been knocked out by the anesthetics.
Therefore, the reason why perceptual learning was not induced in the current study might be because of anesthetic-induced disabling of crucial top-down influences. Importantly, this account would still allow for perceptual priming to occur during anesthesia. For example, the improvement of performance in a word-stem completion task is specific to the “primed” word,5possibly occurring within bottom-up processing55and immune to GA.56,57However, these considerations remain speculations at the moment.
The authors thank Jean-Patrice Gardaz, M.D. (Professor, Department of Anesthesiology, Centre Hospitalier Universitaire Vaudois and University of Lausanne, Lausanne, Switzerland), for help in conceptualization of the study and helpful discussions; Marc Repnow, M.Sc. (Technical Assistant, Laboratory of Psychophysics, Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland), for technical assistance; and Homawoo Holai and Ismail Labgaa (Alternative Service Assistants, Laboratory of Psychophysics, Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne) for helping to collect the data.