Objective. Reflexive responses to pain such as facial reactions become increasingly important for pain assessment among patients with Alzheimer's disease (AD) because self-report capabilities diminish as cognitive abilities decline. Our goal was to study facial expressions of pain in patients with and without AD.
Design. We employed a quasi-experimental design and used the Facial Action Coding System (FACS) to assess reflexive facial responses to noxious stimuli of varied intensity. Two different modalities of stimulation (mechanical and electrical) were employed.
Results. The FACS identified differences in facial expression as a function of level of discomforting stimulation. As expected, there were no significant differences based on disease status (AD vs control group).
Conclusions. This is the first study to discriminate among FACS measures collected during innocuous and graded levels of precisely measured painful stimuli in seniors with (mild) dementia and in healthy control group participants. We conclude that, as hypothesized, FACS can be used for the assessment of evoked pain, regardless of the presence of AD.
Facial Action Coding System
Seniors with moderate to severe dementia tend to present with limitations in verbal communication. As such, the verbal report of pain tends to become less reliable as dementia progresses , and there is a need to focus on alternate measurement methods to improve pain assessment and management among vulnerable dementia patients. Facial expressions of pain have been the focus of several investigators interested in assessing pain among older adults suffering from dementia [2–5].
Presently, the Facial Action Coding System (FACS [6,7]) is the most satisfactory method of objectively describing facial expressions of pain because of the rigorous, explicit criteria of coding pain responsiveness independent of situational demand characteristics . Moreover, the FACS is valid and reliable for monitoring pain-related facial expressions in dementia patients [1,3,9]. To date, FACS has predominately been used to identify the presence or absence of pain [3,10], although one group has reported FACS responses at differing intensities of experimental pain in young adults .
Hadjistavropoulos et al.  examined FACS responses during a graded variety of physiotherapy exercises in older persons with and without dementia. Although more provocative and discomforting physiotherapy activities were accompanied by increased facial activity, Hadjistavropoulos et al.  did not quantify the intensity of pain stimulation, and there has been no systematic investigation of the stimulus–response function of FACS following experimentally controlled levels of noxious stimulation in this population. Hadjistavropoulos et al. also showed that patients with dementia tended to demonstrate somewhat increased facial activity (relative to the control group) in response to movement-exacerbated pain. Others  obtained similar results.
We employed psychophysical techniques to anchor subjective pain report against a stimulus of known intensity and measured FACS responses to these noxious stimuli. Research has shown no difference in the pain threshold of those with mild to moderate AD despite an increase in pain tolerance when compared with age-matched controls [12–14]. However, there are no studies of other points along the stimulus-response function and it remains unclear at precisely what intensity of noxious stimulation differences (if any) start to appear. It has also been suggested that pain unpleasantness in AD may be attenuated more than sensory qualities , but there is relatively little direct evidence to support this view. Our study sought to establish the intensity of noxious stimulation required to elicit a report of just noticeable pain, weak, and moderate pain in those with and without AD and to investigate self-reported unpleasantness ratings and FACS measures in response to these three pain intensities. We studied both electrical and mechanical noxious stimulation, as this allowed us to examine unpleasantness ratings and the properties of FACS across more than one stimulus modality. The following hypotheses were examined:
In both AD patients and controls, we expected increased pain unpleasantness ratings and increased FACS scores in response to increasing stimulus intensities for both types of noxious stimulation (i.e., mechanical and electrical). This hypothesis could hold even if the strength of the relationship between increasing FACS scores and stimulus intensities differs between groups.
Consistent with previous research [12–14], we expected that the stimulus strength required to elicit a report of “just noticeable pain” would not differ between AD patients and age-matched controls, but higher points of the stimulus response function may show group differences [4,9].
All participants signed a consent form prior to participation in the study. The consent form included information regarding all the different processes during participation in this project. The consent form also underscored the fact that participation was voluntary with no adverse consequences to declining participation or discontinuing the study once enrolled. Additional precautions were taken for participants with Alzheimer's disease. Specifically, proxy consent was obtained from the next of kin, the treating physician, and the Victorian Civil and Administrative Tribunal (Melbourne, Australia). This procedure ensured the protection of these potentially vulnerable participants.
Following confirmation of all appropriate institutional ethical clearances, 27 people diagnosed with AD (diagnostic procedures are detailed below) were recruited from the Melbourne Health Cognitive Dementia and Memory Service (CDAMS). The CDAMS is an outpatient service and has strong links with psychogeriatric inpatient services, subacute and long-term care facilities, aged care assessment teams, and support services such as Alzheimer's Australia. It is a multidisciplinary clinic with a staff comprising geriatricians, psychogeriatricians, neuropsychologists, a speech therapist, an occupational therapist, a social worker, and a clinic nurse. In all cases, the medical diagnosis of dementia was made by at least one experienced psychogeriatrician or geriatrician expert in dementia following a physical examination, blood tests, CT and/or MRI scan, informant history, and all other tests outlined by the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA)  task force document. More specifically, the diagnosis of AD was made at CDAMS according to the Diagnostic and Statistical Manual of Mental Disorders IV , the International Classification of Diseases 10 , and the NINCDS-ADRDA  criteria. This approach represents the gold standard for dementia diagnosis in the absence of postmortem results (see, e.g., Tuokko and Hadjistavropoulos ). This complete medical assessment was also used to identify and document medical comorbidities (comorbidities were confirmed by an experienced geriatrician). People with comorbid medical conditions that could influence pain perception (e.g., peripheral neuropathy) or those taking analgesic medications were excluded from the study.
Thirty-six cognitively intact age-matched control participants were recruited from the general community and the volunteer register of the National Aging Research Institute (NARI). NARI is an independent research institute with headquarters in Melbourne and is a center of excellence in Australia for research and training in preventive and public health, clinical, biological and social aspects of aging, and in the evaluation of aged care health service delivery. The NARI volunteer register (from which control volunteers were recruited) represents a database of more than 1,000 older persons who have indicated a willingness to be contacted for participation in research studies of aging and age-related diseases. These volunteers represent a convenience sample, although over half were originally recruited by random sampling of the older persons living in Melbourne. Although a diagnostic examination was not conducted with these volunteer participants, information on demographic characteristics and health (diseases, if any, medications, blood pressure, cognitive status, pain status, falls status, levels of independence in activities of daily living) were recorded following questionnaire/interview at time of entry into the register. This information is updated after each contact with NARI. All control participants were tested with the Mini Mental Status Examination (MMSE)  to ensure that they were cognitively intact (score ≥24). Control volunteers were also screened, via interview, for the presence of comorbid medical conditions that could influence pain perception (e.g., peripheral neuropathy) or for taking analgesic medications, and these people were excluded from the study.
The average age of the participants who received electrical stimulation was 78.6 years (SD = 4.3) for the control group (N = 22) and 78.4 years (SD = 4.0) for the participants with AD (N = 13). For those participants who received mechanical stimulation, the average age of the control group (N = 14) was 78.3 years (SD = 3.8) and 80.0 years (SD = 6.3) for the participants with AD (N = 14). In terms of cognitive status, the control group who received electrical stimulation obtained a mean MMSE score of 28.8 (SD = 2.2) and participants with a clinical diagnosis of AD who received electrical stimulation obtained a mean MMSE score of 21.2 (SD = 5.4). For mechanical stimulation, the control group obtained a mean MMSE score of 29.4 (SD = 0.9) and participants with AD obtained a mean MMSE score of 22.8 (SD = 3.91). Of these participants, 12 control participants and eight participants with AD received both types of stimulation.a
The sensory component of the Gracely Box Ratio Scales for sensory (intensity) and affective (unpleasantness) verbal pain report  was used to determine participants' levels of just noticeable pain (JNP), weak pain, and moderate pain. The self-reports of pain unpleasantness (our outcome measure) were obtained using the unpleasantness component of the Gracely scales. The Gracely Box Ratio Scales have been shown to be valid and reliable . In addition, Chibnall and Tait  have validated the Gracely Box Scales in older persons with cognitive impairments and concluded that they are among the most useful scales. We note that the Gracely pain unpleasantness ratings (as opposed to the pain intensity ratings) represented our dependent self-report variable because the Gracely intensity scale was used to ensure that each participant received the same level of intensity for each intensity level of the stimulus (see the Procedures section). For example, as explained in more detail later, the stimulus intensity for weak pain was set to correspond to a Gracely intensity rating of 3 for all participants. The Gracely pain unpleasantness rating, however, was expected to vary across participants.
The FACS [6,7] is a detailed technical atheoretical tool that objectively classifies facial responses based on the anatomy of the face. Facial movements are broken down into action units (AUs) and a total of 48 have been identified. In addition to the calculation of AU frequencies of occurrence, intensity of each facial response can be categorized using detailed and objective criteria. All coding takes place using slow-action (frame-by-frame) video. In our study, the system was applied by a coder who qualified by passing the examination prepared by the developers of the FACS procedure. As a result of the rigorous explicit criteria employed for FACS coding, the system has been shown repeatedly and consistently to have very high degrees of inter-rater and intra-rater reliability (see, e.g., Craig et al.  for a review).
For this study, an intracoder reliability assessment method was used whereby the coder was given a random selection constituting 15% of the total coded data. Then the coder, without having access to the original scores for those selected pain segments, recoded this randomly selected 15%. The reliability coefficient (calculated using the conservative agreement method described by Ekman et al. ) indicated very good reliability for AU frequency (0.82). To assess intra-rater reliability with respect to AU intensity, an intraclass correlation was calculated (0.83).
The noxious stimulus was administered using either an electrical–thermal pain stimulator or a mechanical stimulator device. The electrical–thermal pain stimulator was designed and constructed by Gorman ProMed (Carnegie, Victoria, Australia). Electrical stimulation was administered through electrodes placed 3 cm apart on the skin on the retromalleolar aspect of the ankle, over the likely course of the sural nerve of the left ankle. Each stimulus lasted for 5 seconds, containing 5 × 200 millisecond pulses at a frequency of 250 Hz. Threshold values were measured in milliamperes.
The mechanical stimulation device [23,24] was a pressure stimulator that is hydraulically driven and uses forces generated by static weights to transmit pressure to the thumbnail of the participant. The appeal of thumbnail pressure is that the nail bed is quite sensitive and pain is experienced at pressures that are well below the intensities that cause any tissue damage. Lead-filled containers (which range in weight from 0.25 to 5.0 kg) are placed on one of two wooden platforms that each rest on the piston of a standard 10-mL plastic syringe. Both platform syringes connect to a 10-mL stimulating syringe and a 10-mL manual-control syringe via water-filled plastic extension tubing. The piston on the stimulating syringe is mounted with a circular rubber stopper (1-cm diameter) and is held in position above the participant's thumbnail by a plastic T-joint pipe (3.5-cm diameter). When weight is placed on one of the platforms, the piston of the platform syringe descends, transmitting pressure through the water-filled plastic tubing to the participant's thumbnail via the rubber-tipped piston of the stimulation syringe. Evaluation of responses to the mechanical pain stimulator in the psychophysical lab and during functional brain imaging sequences has established the reliability and utility of the device . In addition, the same device has recently been utilized in additional research without any adverse effects .
After participants were recruited and AD diagnosis confirmed (see above for a detailed discussion of diagnostic procedures), pain thresholds were determined using a multiple random staircase method . In this procedure, one staircase is randomly chosen for each trial, and the response from the participant determines the intensity that will next be delivered for that staircase. For example, if a stimulation is delivered from staircase 1 (sensation threshold) and the participant does not feel anything, the following stimulation for that staircase will be of increased intensity, and vice versa. For the purposes of this study, the turning points for staircases 3 and 4 (weak pain and moderate pain) were set at ratings of 8 and 11, respectively, from the Gracely Box Scale for intensity. This multiple random staircase method reduces the effect of bias that can result from participants becoming aware that their response to each stimulation determines the intensity of subsequent stimulations . While it is not possible totally to eliminate the possibility of summation effects from repeated pain stimuli, the final stimulus protocol used with electrical stimulation was the product of several hours of testing aimed at reducing such effects. One of the known characteristics of the mechanical stimulator is that, at the levels employed in this study, summation effects are minimal. In the context of the multiple staircase procedure, there were relatively few stimuli delivered at the (self-rated) “moderate” level of intensity and, despite the randomized stimulus delivery, participants rarely received three or more consecutive stimuli at this level. The staircase would also track downward if the person was experiencing any sensitization.
Participants were asked to describe the type of sensation that they experienced using the intensity component of the numeric rating scale (NRS) in order to determine the level of stimulation needed to induce JNP threshold, weak pain, and moderate pain for each participant. A total of approximately 70 stimuli were delivered during the staircase procedure using electrical stimulation and 45 stimuli using mechanical stimulation. Thresholds for the JNP, weak pain, and moderate pain were derived from the mean of the last five responses in each staircase. Ratings of pain affect were derived using the unpleasantness component of the NRS and the method of constant stimuli model, in which participants received a randomized sequence of stimuli at their individually derived threshold values (JNP, weak, moderate) each presented on a total of three occasions. This same randomized sequence was subsequently presented a second time during which time video recordings of each participant's facial expressions were made for each level of noxious stimulation as well as during a calm, stimulus-free baseline period. We note that the first administration of the randomized sequence was not videotaped because it was felt that the patients' verbal responses, provided during the first administration (often while the stimulus was still active), would interfere with the accurate coding of facial expressions of pain. It is important to note that, although the research staff did not have information on dementia diagnosis when testing pain thresholds, it is impossible to remain blind to the participants' cognitive status during the test session (i.e., those with moderate cognitive impairment are clearly recognized by task performance during testing). However, we do not believe that this knowledge would compromise the psychophysical results because it is the participants themselves who determine the thresholds for just noticeable, weak, and moderate pain by their verbal responses to presented stimuli. Each participant remains blind to stimulus intensity at all times and the use of multiple random staircase test procedures ensures that there is no contingency between the verbal response and the intensity of the next stimulus delivered. Each of the three staircases (JNP, weak, moderate pain) are presented in a random order and have an automatic computer-generated algorithm to determine the next stimulus intensity for that staircase based on the previous verbal response. As a result, we do not believe that it is likely that research staff could directly influence the threshold value.
In terms of the videotaping, video recording equipment was placed so that the camera was in direct line with the face, and the image consisted of the front of the face including the lower extremity of the chin to the start of the hairline at the forehead. Recordings were made for each of the three levels of sensation, as well as a prestimulus period, 5 seconds before the stimulus onset for each sensation. In addition, participants were filmed when they were not receiving the stimulus in order to make comparisons between the facial expressions when relaxed (i.e., baseline) and when experiencing pain (i.e., JNP, weak pain, and moderate pain). To facilitate the editing procedure, the recorded segments were marked with a numerical indicator between each stimulus administration. We used the Pinnacle DV500 DVD (Pinnacle Systems, Mountain View, CA) and Adobe Premier 6.5 software (Adobe Systems Inc., San Jose, CA) to view the video recordings frame by frame. The segments (i.e., baseline and noxious stimulation segments) were randomized within participants and across groups by someone other than the FACS coder. All additional footage (e.g., talking between stimulus administrations) was deleted. Each individual segment was separated from the next by white numbers on a black background indicating the end of that particular segment (e.g., 1, 2, 3, etc.). This ensured that the FACS coder (who passed the required FACS reliability examination administered by the developers of the system) was unaware whether she was watching a JNP, weak pain, or moderate pain segment (only the baseline segment was identified). We also note that all FACS coding was undertaken blind to participant diagnosis and to the intensity of noxious stimulation, and this represents a major strength of the current study.
Self-Report Ratings of Pain
According to Hypothesis 1, self-report ratings of pain (i.e., pain unpleasantness) were expected to increase as a function of stimulus intensity for both types of stimulation. Consistent with the hypothesis, we observed that the mean pain unpleasantness ratings, as measured by the unpleasantness component of the Gracely scales, increased as a function of sequential threshold levels, for both electrical, F(2,66) = 166.70, P < 0.001, and mechanical, F(2,52) = 116.04, P < 0.001, stimulation (see Figure 1). No significant group or interaction effects were found. For electrical stimulation, planned comparisons using the least significant difference (LSD) method  showed significant differences between JNP and weak pain, t(34) = 11.20, P < 0.001, JNP and moderate pain, t(34) = 16.63, P < 0.001, as well as weak and moderate pain, t(34) = 10.92, P < 0.001. For mechanical stimulation, planned comparisons showed that significant differences existed between JNP and weak pain, t(27) = 10.11, P < 0.001, JNP and moderate pain, t(27) = 12.54, P < 0.001, as well as weak pain and moderate pain, t(27) = 6.64, P < 0.001.
Mean Gracely Box Affective Scale ratings for each type of noxious stimulation (i.e., electrical stimulation and mechanical stimulation) broken down by group (i.e., Alzheimer's disease [AD] and control). JNP = just noticeable pain.
Before testing the aspect of Hypothesis 1 relating to FACS (i.e., that FACS scores would increase as a function of pain stimulation), we examined the frequencies of AU activation during painful stimulation. AUs that were activated during painful stimulation among at least 25% of participants (presented separately for the electrical and/or mechanical stimulation) are shown in Table 1 Consistent with prior research among seniors, it appears that it is the overall facial activity and not individual AUs that differentiate between painful and nonpainful periods . This conclusion is based on the observation that there is no AU that manifests consistently across patients during all of the pain segments in this study. Therefore, we focused our main analyses on overall facial activity (i.e., total frequency and average intensity of all AUs present). We do note, however, that two AUs occurred with high frequency during the pain segments. These AUs were AU-43 (eye closure), and AU-45 (blinking).b Overall facial activity variables are summarized in Table 2.
Means and standard deviations of participants' action unit frequency and intensity for mechanical and electrical stimulation by group
Just noticeable pain
Just noticeable pain
According to Hypothesis 1, FACS scores (i.e., both frequency and intensity of facial responses) were expected to increase as painful stimulus intensities increased. We first conducted a 2 × 4 mixed model (group by level of stimulation) anova with FACS frequency as the dependent variable. Results of the anova indicated that, for electrical stimulation, there was a significant within-subjects main effect with respect to total AU frequency, F(3,99) = 20.39, P < 0.001. As Figure 2 indicates, AU frequency scores increased as a function of stimulus intensity and there were no significant group or interaction effects. Planned comparisons, using the LSD method , showed significant differences between baseline and JNP, t(34) = 4.26, P < 0.001, baseline and weak pain, t(34) = 6.18, P < 0.001, as well as baseline and moderate pain, t(34) = 6.84, P < 0.001. In addition, significant differences were found between JNP and moderate pain, t(34) = 2.09, P < 0.05, and weak and moderate pain, t(34) = 2.37, P < 0.05. However, no significant differences were found between JNP and weak pain.
Facial Action Coding System total frequency scores for both types of noxious stimulation (i.e., mechanical stimulation and electrical stimulation) broken down by group (i.e., Alzheimer's disease [AD] and control). AU = action unit; JNP = just noticeable pain.
To test whether FACS intensity scores also changed as hypothesized (Hypothesis 1), we conducted a second 2 × 4 mixed model (group by level of stimulation) anova with FACS intensity scores as the dependent variable. Results showed that, consistent with the our hypothesis, the increases in FACS intensity are significant F(3,99) = 20.81, P < 0.001. As Figure 3 illustrates, FACS intensity scores increased as a function of noxious stimulation. Similar to total AU frequency, no significant interaction or group effects were found. Planned comparisons using the LSD method showed that FACS intensity scores differed significantly between baseline and JNP, t(34) = 6.80, P < 0.001, baseline and weak pain, t(34) = 7.09, P < 0.001, and finally between baseline and moderate pain, t(34) = 6.26, P < 0.001. Significant differences were also found between JNP and moderate pain, t(34) = 2.05, P < 0.05. However, there were no significant differences between JNP and weak pain, or weak pain and moderate pain.
Facial Action Coding System average intensity scores for both types of noxious stimulation (i.e., mechanical stimulation and electrical stimulation) broken down by group (i.e., Alzheimer's disease [AD] and control). AU = action unit; JNP = just noticeable pain. AU intensity ratings correspond to Ekman and Friesen's a-b-c-d-e scale as follows: a = 1, b = 2, c = 3, d = 4, e = 5.
Following the tests of Hypothesis 1 (as it pertains to FACS scores) for participants who received electrical stimulation, we proceeded to test the same hypothesis for participants who received mechanical stimulation. We conducted a 2 × 4 (group by level of stimulation) mixed model anova using FACS frequency scores as a dependent variable and another 2 × 4 mixed model anova using FACS intensity scores as a dependent variable. As was the case with the analysis of electrical stimulation FACS scores, a within-subjects effect was identified for total AU frequency, F(3,78) = 7.53, P < 0.001, with scores increasing as a function of stimulus intensity. There were no significant between-group or interaction effects. We followed up this within-subjects effect with planned comparisons using the LSD method. Significant differences were found between baseline and JNP, t(27) = 4.87, P < 0.001, baseline and weak pain, t(27) = 5.13, P < 0.001, as well as baseline and moderate pain, t(27) = 3.52, P < 0.01.
In terms of FACS intensity ratings, as expected, we found a significant within-subjects main effect (but no interaction or group effect) F(3,78) = 9.51, P < 0.001 (see Figure 3), indicating that scores increased as a function of stimulus intensity. Planned comparisons showed significant differences between baseline and JNP, t(27) = 4.09, P < 0.001, baseline and weak pain, t(27) = 4.12, P < 0.001, and baseline and moderate pain, t(27) = 4.05, P < 0.001. Additional differences were found between JNP and moderate pain, t(27) = 2.37, P < 0.05 as well as weak pain and moderate pain, t(27) = 2.46, P < 0.05.
In order to explore the possibility that there may be a between-group difference with respect to the strength required to elicit a response of JNP, weak pain, and moderate pain (see Hypothesis 2), we compared the groups with respect to stimulus strength (measured in milliamperes for electrical stimulation and kilograms for mechanical stimulation—see Table 3) using 2 × 3 (group by stimulus strength needed to elicit JNP, weak, and moderate pain) anovas. A separate anova was conducted for each type of stimulation. There were no significant group main or interaction effects, although as expected, the stimulus strength was significantly higher (as a function of study design) in order to elicit higher ratings for both electrical F(2,66) = 38.23, P < 0.000 and mechanical stimulation, F(2,52) = 167.25, P < 0.000 (see Table 3).
Means and standard deviations of participants' threshold values for mechanical and electrical stimulation
Mean mechanical threshold (kg)
Just noticeable pain
Mean electrical threshold (milliamperes)
Just noticeable pain
Additional Exploratory Analyses
Electrical vs Mechanical Stimulation for FACS Measures
We conducted an exploratory analysis to determine whether there were differences in FACS scores as a function of type of painful stimulation. Specifically, a subgroup of 20 participants received both types of noxious stimulation during two separate sessions. To determine if any differences existed between electrical and mechanical stimulation in terms of the FACS indices, a series of 2 × 2 (group × type of stimulation) anovas were conducted. Significant differences in total AU frequency were found between the two types of stimuli (i.e., mechanical and electrical) at all levels of pain stimulation: JNP, F(1,18) = 47.7, P < 0.001, weak pain F(1,18) = 27.0, P < 0.001, and moderate pain, F(1,18) = 12.6, P < 0.01. In all instances, electrical stimulation frequency scores were higher than mechanical stimulation scores. No group effects or interactions were found at any level of stimulation.
Significant differences in average AU intensity were only found at the two levels of stimulation, JNP, F(1,18) = 8.0, P < 0.001, and weak pain, F(1,18) = 5.2, P < 0.05. In both cases, electrical stimulation intensity scores were higher than mechanical stimulation scores. Similar to the frequency analysis, no group effects or interactions were found.
The Relationship of Self-Report with FACS
In order to explore the relationship of FACS scores with self-report, we calculated correlations of pain unpleasantness ratings with the FACS indices (i.e., frequency and intensity) for both types of stimulation (separately for each group, type of stimulus, and level of stimulation). Due to the high number of correlations conducted (i.e., 32), a conservative critical value of 0.001 was used. None of these correlations were significant. That is, significant correlations were not found regardless of group membership (i.e., control or AD) and type of stimulation (i.e., electrical or mechanical stimulation).
This is the first study of older adults with and without AD that involved the investigation of specific, accurately measured levels of painful stimulation and the assessment of corresponding facial responses. Moreover, we investigated the relationship between self-report ratings of pain unpleasantness and FACS. We demonstrated the utility of FACS in the pain assessment of seniors regardless of the presence of AD using two different stimulus modalities (i.e., electrical and mechanical). Our findings are consistent with the idea that AD does not attenuate pain reactivity up to moderate levels of pain.
Previous research involving the FACS and participants with and without dementia involved pain in clinical settings [3,9], and the level of painful stimulation was not varied systematically using precise indices of stimulus intensity. In this study, psychophysical techniques were used to ensure that the precise levels of stimulation were known. Given the subjective nature of pain, this procedure allowed for comparisons of facial expressions and unpleasantness ratings at specific levels of pain stimulation (i.e., JNP, weak pain, and moderate pain). Most studies have investigated facial responses to pain at only one level of painful stimulation [1,3,8,10,26], rather than a quantitative evaluation of FACS over the stimulus-response range of noxious stimulation. This is an important contribution of the current study because FACS becomes a potentially more useful tool if it can be shown to index increasing levels of pain intensity in addition to merely distinguishing between the presence or absence of pain.
Self-Report of Pain
Consistent with Hypothesis 1, as the stimulus intensity increased so did the pain unpleasantness ratings. These increases were found regardless of the presence or absence of AD. Therefore, consistent with previous studies (see, e.g., Hadjistavropoulos ), these findings provide further convincing evidence that participants with mild to moderate dementia can provide valid self-report ratings of pain.
FACS generally identified differences in facial expression associated with different levels of painful stimulation. In addition, these differentiations were found regardless of group membership, which is consistent with the findings of other researchers .
As shown in Table 1, certain AUs occurred with high frequency (i.e., displayed among at least 25% of participants) during painful levels of stimulation. These AUs are similar to those identified in previous studies involving younger adults [8,29]. We note that blinking and eye closure occurred at especially high frequencies (although blinking was also elevated during baseline). Nonetheless, no specific AUs universally characterized facial expressions of pain among our participants. The identification of AUs that occur commonly during pain states (perhaps at levels of stimulus intensity that are higher than the levels used in this study) may have potential in helping distinguish pain from other distressing states among seniors with severe dementia. We must note that the stimulus-response function of FACS is not linear (see Figures 2 and 3). Indeed, with noxious electrical stimulation, the FACS stimulus-response function almost looks like a dichotomous response pattern. That is, there is a huge increase in FACS activity between the “no pain” and “pain” test conditions, but little further increase in FACS responses with successively higher levels of pain. Mechanical stimulation looks more linear, with a relatively uniform and gradual increase in FACS accompanying each progressive increase in pain intensity. Further work is needed to better characterize the stimulus-response function of FACS in response to different modalities and durations of noxious stimulation.
The Relationship of Self-Report and FACS Scores
Contrary to expectations, no significant correlations were found between the FACS measures and pain unpleasantness ratings. Conclusions concerning the relationship of self-report and behavioral ratings (i.e., the FACS) are inconsistent in the literature (see, e.g., Labus et al. ). As in the present study, others have failed to detect significant relationships between observational indices of pain and self-report measures [11,30]. It appears that FACS measures a different aspect of the pain experience than does self-report. That is, the FACS probably measures more reflexive aspects of pain (i.e., automatic facial responses to noxious stimulation) that are less subject to the influence of cognitive and social factors , whereas self-report reflects more abstract, higher levels of processing and is more susceptible to personal and social biases, such as “downplaying” the pain experienced . It could also be that FACS would correlate more highly with self-report of pain intensity than with self-report of pain unpleasantness.
The absence of a significant correlation between facial activity and self-report scores suggests that the two types of pain expression may be serving different functions from a socioevolutionary standpoint (i.e., the facial response could be considered to represent an immediate signal of danger, whereas verbal communications are usually intended to provide more detail and specific information to potential caregivers and bystanders). Moreover, as stated earlier, FACS (compared with self-report) likely measures more automatic aspects of the pain experience.
Electrical vs Mechanical Stimulation
When comparing electrical and mechanical stimulation across participants, analysis revealed that the frequency of facial actions during electrical stimulation was considerably greater than that involved in mechanical stimulation. This difference was found for both participants with and without AD. One possible explanation for this finding is that electrical stimulation occurs more suddenly than mechanical stimulation, which intensifies somewhat more gradually over a prolonged period of time. In addition, the nature of electrical stimulation is more sharp and sudden, possibly creating a startle response , whereas mechanical stimulation involves a dull sensation of pain. These findings are consistent with previous research showing that electrical stimulation is more likely than other types of stimulation to produce a response that tends to lead to a higher AU frequency .
An interesting pattern emerged with FACS intensity. Although there were significant differences between the two types of stimulation at JNP and weak pain, this difference did not exist for moderate pain. This finding suggests that, although these two types of stimulation differ in response at very low levels of pain, as the intensity of the pain stimulus increases, facial responsiveness becomes comparable, regardless of the type of pain stimulation administered. One possible explanation for this finding is that, at higher levels of noxious stimulation, a greater majority of the intensity of the AUs expressed is due to pain rather than a startle response that is often associated with electrical stimulation.
As expected, no significant differences were found between participants with and without AD with respect to the level of stimulation needed to elicit noticeable facial responses to pain. This finding is consistent with previous research that examined pain thresholds using self-report ratings [12–14]. Collectively, these results suggest that sensitivity to weak to moderate pain is not altered in the early stage of AD.
Our findings suggest that at threshold to moderate levels of pain, the presence of early AD does not have a major impact on a patient's nonverbal expressions nor on ability to use a self-report measure to indicate pain, regardless of the type of noxious stimulation (i.e., electrical or mechanical stimulation). In addition, FACS was shown to be capable of generally differentiating among very specific and low levels of stimulus intensity, supporting the sensitivity of this measure. This is an important finding, considering the lack of differences based on group membership (e.g., AD vs control group), because it suggests that it is feasible to assess pain without relying solely on self-report, which is known to become less reliable with cognitive decline (see Hadjistavropoulos  for a review). Moreover, this result is consistent with previous research [1,3]. Nonetheless, we acknowledge that our sample size sets some limits on the generalizability of our conclusions. It would be important for future research to replicate our findings using larger samples. With respect to our use of FACS, we acknowledge that, although the inter-rater reliability of the system has been demonstrated repeatedly for formally qualified coders , we were only able to provider intra-rater (as opposed to inter-rater) reliability information.
While FACS is too time-consuming for use in busy clinical settings, a potential clinical application would involve the development and validation of simpler systems designed to assess facial actions. Such systems may be suitable for use in clinical settings. Nonetheless, our results support the great potential of FACS as a research outcome measure in treatment studies involving seniors with and without dementia. Specifically, while previous research has shown that FACS scores correlate with global ratings of pain based on facial expressions , it has also been demonstrated that global ratings, based on facial expressions, are subject to observer bias. For example, nurses give lower global ratings of pain than university students observing the same patients , and patient physical attractiveness and gender  bias global ratings of pain. As such, in clinical research contexts, where precision and objectivity are required, FACS (with its atheoretical orientation) could minimize the effects of subjective judgements. Moreover, at higher pain intensities than were involved in our study, it is possible that specific and unique facial displays of pain  would be easier to identify through the use of FACS than through subjective observations of facial activity.
This study was supported, in part by a Canadian Institutes of Health Research Investigator Award to Thomas Hadjistavropoulos, as well as a Social Sciences and Humanities Research Council of Canada Master's Award and a University of Regina Center on Aging and Health Graduate Fellowship to Amanda Lints-Martindale. It was also supported by The National Health and Medical Research Council of Australia (Grant no. 219291).
↵a The original intent of this study was to use only electrical stimulation. However, there was also a plan for a second but independent study that would involve the administration of discomforting stimulation inside an MRI scanner. The level of electrical stimulation inside the scanner was to be determined through the pretesting that is described in this article. However, after the present study commenced, it became known (from independent research) that electrical stimulation could potentially cause problems with the MRI scanner (occasional skin burns surrounding the metal stimulation electrodes were reported under certain test conditions). Due to these safety concerns, participants who had received electrical stimulation were invited to an additional session where mechanical stimulation was delivered to determine thresholds for use during fMRI sessions (and thus they received both electrical and mechanical stimulation). Subsequent participants received mechanical stimulation. It should be noted there were no negative incidents or complications during electrical stimulation among the participants during the psychophysical testing. Nonetheless, this situation allowed us to compare responses to electrical vs mechanical stimulation within participants.
↵b A separate series of analyses focusing on the combination of these two AUs (43 and 45) showed that discrimination across levels of stimulation was not as effective as when overall facial activity was considered.
. An examination of pain perception and cerebral event-related potentials following carbon dioxide laser stimulation in patients with Alzheimer's disease and age-matched control volunteers. Pain Res Manag 2001;6(3):126–32.
Clinical diagnosis of Alzheimer's disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology 1984;34(7):939–44.