Introduction

It has been argued that all behaviours are governed by motivational circuits that are cued by either appetitive or aversive stimuli (Bonnet et al. 1995; Lang et al. 2000). Aversive stimuli are characterized as unpleasant and highly arousing. An animal’s response to aversive stimuli depends on stimulus proximity and intensity; animals will respond with defensive aggression to close, intense aversive stimuli; and ‘postural freezing’ (i.e. immobility) to distal or less intense stimuli (Amorapanth et al. 1999; Blanchard et al. 1986; Lang et al. 2000).

There is some evidence to suggest that humans also demonstrate ‘postural freezing’ behaviours to aversive or threatening affective picture stimuli, based on subtle changes to postural control (Azevedo et al. 2005; Facchinetti et al. 2006). Studies that have contrasted unpleasant and pleasant affective pictures have demonstrated an effect of unpleasant valence on postural control during two-legged stance. However, the direction of, and variables involved in, this effect have been somewhat inconsistent across studies. For instance, Hillman et al. (2004) demonstrated that only women leaned significantly backwards while watching unpleasant pictures compared to pleasant pictures. In contrast, Stins and Beek (2007) found that subjects of both genders leaned slightly forward in response to unpleasant pictures compared to pleasant ones, although neutral and unpleasant pictures induced a comparable effect. Likewise, Azevedo et al. (2005) demonstrated increased centre of pressure (COP) frequency and decreased amplitude in the medial–lateral (ML) plane in response to unpleasant pictures, compared to pleasant ones or neutral ones; yet Stins and Beek (2007), who used similar picture groups, did not see any significant change in ML sway amplitude during two-legged stance. In the anterior–posterior (AP) plane, Roelofs et al. (2010) demonstrated reduced COP amplitude to unpleasant pictures (angry faces), compared to pleasant ones (happy faces); and yet neither Azevedo et al. (2005) nor Facchinetti et al. (2006) found an effect of unpleasant pictures on COP amplitude in the AP plane. In sum, there is ample evidence to suggest that there is an effect of unpleasant pictures on postural control, yet the true extent of this effect is unclear.

The inconsistencies across these studies may be attributed to either methodological limitations concerning posturographic measures and/or a lack of control or consideration for the potential effects of arousal. For example, some studies (Hillman et al. 2004; Roelofs et al. 2010; Stins and Beek 2007) used sampling durations too short (<10 s) to accurately characterize the full spectrum of postural sway (Carpenter et al. 2001; van der Kooij et al. 2011). As a result, the short samples lead to under-representation of the lowest frequency components of the COP signal (thought to represent movement of the body) and over-representation of the higher-frequency oscillations of forces acting under the feet (thought to represent regulation of postural muscle tone; Winter et al. 1998). As such, it is not clear whether the unpleasant valence effects reported in these studies reflect sustained changes in whole-body leaning; or short duration, transient responses to the pictures.

Furthermore, despite the theoretical link between valence and arousal in emotion, none of the aforementioned studies have systematically manipulated arousal, independently from valence, to determine its effects on postural control. Some studies have grouped arousing and calming pictures together (Hillman et al. 2004; Stins and Beek 2007), others kept arousal constant while manipulating valence to study across valence conditions (Azevedo et al. 2005), and the remainder did not attempt to control arousal at all (Facchinetti et al. 2006; Roelofs et al. 2010). This is a very important limitation, because (a) there is evidence from other paradigms indicating that arousal alone may be associated with some postural changes (Maki and McIlroy 1996); (b) postural control and arousal both change in parallel when presented with a postural threat (Brown et al. 2006; Carpenter et al. 2006; Davis et al. 2009; Huffman et al. 2009); (c) transcranial magnetic stimulation studies suggest that corticospinal tract excitability is more heavily linked to arousal than valence (Baumgartner et al. 2007; Coombes et al. 2009); and finally, (d) the effects of valence on stretch reflexes in a postural muscle (soleus) have been shown to be dependent on arousal when both valence and arousal are manipulated independently (Bonnet et al. 1995). In particular, while stretch reflexes were significantly larger when subjects watched unpleasant pictures compared to pleasant ones in the low-arousal condition, there was no effect of valence in the high-arousal conditions. As such, there is considerable evidence to suggest that arousal may play a significant role in shaping postural responses to emotional stimuli.

Therefore, the aim of this study was to manipulate both arousal and valence, separately, to determine their potential independent and interactive effects on postural control. We used longer sampling durations to capture the full spectrum of postural sway characteristics. Based on the available literature concerning the effects of affective pictures on postural control, we predicted significant main effects between unpleasant and pleasant valence on AP COP mean position, frequency and amplitude, as well as ML frequency and amplitude. However, based on the work by Bonnet et al. (1995), we further predicted that the effects of valence would be dependent on arousal, such that significant differences between valence conditions would be apparent when arousal was low, and yet unpleasant and pleasant valence would be comparable when arousal was high.

Methods

Participants

Fifty-two subjects (mean age 24.3, range: 19–32 years; 26 women) participated in this study. Participants were excluded from the study if they had a known neurological, orthopaedic, vestibular, or uncorrected visual impairment that may have impeded them from standing quietly or viewing the picture stimuli. All participants were provided written informed consent, and the study was approved by the University of British Columbia Clinical Research Ethics Board.

Materials and procedure

Manipulation of valence and arousal

Affective pictures from the International Affective Picture System (IAPS) (Lang et al. 2008) were used to modify emotional valence and arousal. Four experimental picture groups consisting of 15 pictures each were assembled from the IAPS directory. Normative Self-Assessment Manikin (SAM) (Bradley and Lang 1994) values compiled through IAPS were used to divide pictures into four distinct groups: Unpleasant low-arousal, unpleasant high-arousal, pleasant low-arousal, and pleasant high-arousal.Footnote 1 The normative scores for each block are graphed in Fig. 1, clearly demonstrating independence of groups.

Fig. 1
figure 1

Shaded boxes represent the normative valence and arousal ranges of the IAPS pictures used in the four experimental picture blocks. The overlaying points (women: circles; men: squares) represent the measured mean valence and arousal ratings for each of the experimental picture groups; error bars represent standard error measure

Full size image

Procedure

Participants performed a total of five standing trials. A common first trial with neutral valence and low-arousal picturesFootnote 2 was used to mitigate potential first trial effects (Adkin et al. 2000); this trial was not further analysed. After the neutral trial, participants performed each of the four experimental trials that were counterbalanced across participants to minimize potential order effects. In each trial, participants were required to stand quietly for 90 s with their feet side-by-side, approximately shoulder-width apart with arms hanging freely and eyes fixated on a 17-inch flat screen computer monitor adjusted to standing eye line located 130 cm in front of the subject. Foot position was marked on the forceplate to ensure consistent positioning across trials. A 2 min seated break was given between each trial, and subjects were required to stand in place for at least 30 s prior to initiation of each trial to allow autonomic responses associated with standing up (Olufsen et al. 2008) and transient COP components to stabilize (Carroll and Freedman 1993).

Each trial consisted of fifteen pictures from the same experimental picture group displayed for 6 s each for a continuous 90 s picture-viewing period. Participants were given specific instructions to watch the pictures, not look or turn away from the screen, and not speak during the trial.

Dependent measures

Posturography

Ground reaction forces and moments of force were measured with a forceplate (#K00407, Bertec, USA) and sampled at 100 Hz (Power 1401, CED, UK). The signal was low-pass filtered at 5 Hz with a 2nd order Butterworth filter, and COP was calculated over the 90 s picture-viewing period for each trial in both the AP and ML directions. From each trial, the AP COP mean position was calculated from the filtered signal; subsequently, the mean power frequency (MPF) and root mean square (RMS) from both the AP and ML directions were calculated from the unbiased COP signal.

Physiological arousal

Electrodermal activity (EDA) was measured from the thenar and hypothenar eminences of the non-dominant hand (model 2502, CED, UK). The EDA signal was sampled at 1,000 Hz and low-pass filtered at 5 Hz offline. The EDA signal was baseline corrected by subtracting the mean value of a 20 s period preceding picture onset then averaged over the same 90 s picture-viewing period from which posturographic measures were calculated.

SAM ratings

After each standing trial, subjects were asked to rate the picture block as a whole with the SAM valence and arousal scales. The SAM scales were designed to measure how pleasant and how emotionally arousing a stimulus is on a scale of 1–9 (Bradley and Lang 1994). Low scores indicate unpleasant or non-arousing pictures, and high scores indicate pleasant or arousing pictures (Fig. 1). For example, an extremely pleasant yet very calming picture block would have a valence score of 9 and an arousal score of 1.

Statistical analyses

Since there is evidence that men and women have different affective responses to IAPS pictures (Lang et al. 2008) and that these gender differences could potentially translate to different postural responses to these pictures (Hillman et al. 2004), we included gender as a between-subjects factor in our statistical analysis. Two (valence: low, high) by two (arousal: low, high) by two (gender: female, male) mixed model ANOVAs were used to test for significant differences across conditions for each dependent variable. Any interactions were explored post hoc with paired samples t tests. The criterion for statistical significance was set to α = 0.05 for all statistical tests, and effect sizes are reported as partial eta squared (ηp²); the results from all statistical tests are reported in Table 1.

Table 1 Summary of results from statistical tests for effects of valence (V), arousal (A), and gender (G), as well as their interactions; statistically significant effects are bolded and italicized
Full size table

Results

Subject exclusions

Two subjects (one woman) were excluded from the study because their COP signals were dominated by high-frequency components that were too high to have occurred in normal quiet sway (Carpenter et al. 2001) and were therefore considered to contain non-biological artefacts. As such, the total number of subjects included in the final analysis was 50 (25 women). Due to equipment malfunction, EDA data from two further subjects were not analysed, though these subjects were included in the remainder of the tests.

SAM ratings and physiological arousal

There was a significant arousal by gender interaction (P = 0.004) on rated SAM valence (Fig. 1). The difference in rated valence from low- to high-arousal picture conditions was greater for women than it was for men (t 48 = 2.989, P = 0.004, η² = 0.157). There were also significant main effects of valence (P < 0.001) and arousal (P < 0.001) on rated SAM valence. Higher (more pleasant) valence ratings were observed in the pleasant compared to unpleasant conditions and in the low compared to high-arousal conditions. Rated SAM arousal was significantly influenced by a valence by arousal interaction (P < 0.001). SAM arousal was consistently rated higher in the high compared to low-arousal conditions; however, the difference between arousal levels was greater for the pleasant picture groups (t 49 = −13.515, P < 0.001, η² = 0.788) than for the unpleasant ones (t 49 = −11.478, P < 0.001, η² = 0.729; Fig. 1). Note that this distinction across valence conditions coincides with the normative IAPS ratings of the pictures used in this study. There were no significant effects of gender on rated SAM arousal. Finally, there was a significant increase in EDA levels in the high compared to low-arousal conditions (P < 0.001; Fig. 2). There were no significant interactions or main effects of valence or gender on EDA.

Fig. 2
figure 2

Main effects of arousal and valence are demonstrated for EDA and COP-dependent measures with mean (±SE) change scores. Arousal effects were calculated as high-arousal minus low-arousal conditions (pooled across valence); positive values indicate larger values in the high-arousal conditions. Valence effects were calculated as pleasant minus unpleasant conditions (pooled across arousal); positive values indicate larger values in the pleasant conditions and negative values indicate larger scores in the unpleasant conditions. Note: Asterisks indicate statistical significance at P > 0.05; also, for mean position, larger scores indicate a backward lean

Full size image

Posturographic measures

There was a statistically significant main effect of arousal condition on AP MPF (P = 0.005). AP MPF was significantly higher in the high compared to low-arousal conditions (Fig. 2) Although AP RMS was on average lower in the high compared to low-arousal conditions, neither AP RMS nor AP mean position was significantly influenced by main effects of arousal. There were no significant main or interaction effects involving valence or gender on any of the AP postural measures. While ML MPF was on average higher in the high compared to low-arousal conditions and also in the unpleasant compared to pleasant conditions (Fig. 2), there were no statistically significant main effects or interactions of arousal, valence, or gender on either ML MPF or RMS.

Discussion

Contrary to our original hypotheses, we did not observe any significant effects of valence or interaction effects between valence and arousal on postural control. The lack of agreement between the results of the current and prior studies in the ML plane may be attributed to methodological differences in the studies; subjects in the current study stood with their feet shoulder-width apart, whereas prior studies demonstrating an effect of unpleasant valence in the ML plane had subjects stand with their feet together (Azevedo et al. 2005; Facchinetti et al. 2006), or on one-leg only (Stins and Beek 2007). As such, our subjects, with their wide base-of-support, may have been too stable in the ML plane to demonstrate the characteristic postural responses to unpleasant pictures that have been previously observed with narrower, less stable, stance widths. Within the AP plane, where stance position was similar across all studies, the absence of valence effects on MPF and RMS amplitudes observed in the current study was in agreement with findings of Azevedo et al. (2005) and Facchinetti et al. (2006). While Roelofs et al. (2010) found decreased amplitudes with unpleasant pictures, the relatively short sampling duration used to calculate these measures (3 s) may have confounded these results and makes it difficult to compare with the other studies that have used longer sampling durations (Carpenter et al. 2001; Carroll and Freedman 1993; van der Kooij et al. 2011).

Previous studies have also suggested that unpleasant pictures induce a leaning effect in the AP plane; however, we cannot confirm this effect with these data. As discussed in the introduction, the studies that have demonstrated a leaning response to unpleasant pictures were not in agreement as to the direction of the effect. Stins and Beek (2007) observed forward leaning in response to unpleasant pictures, whereas Hillman et al. (2004) found that women leaned backwards. However, Hillman et al. (2004) also found that men tended to lean forward with unpleasant pictures, yet this effect was not statistically significant. Since we sampled equally across genders, it was possible for us to test the possibility that the direction of lean is dictated by gender. However, despite the fact that there were slight differences in how pleasantly men and women viewed the pictures, we found no main or interactive effects of gender on AP mean position, or any other postural measure. Therefore, we conclude that there was no significant leaning effect induced by the pictures used in this study. The difference between this and previous studies that have found a leaning effect might be attributed to their reliance on very short sampling durations (1 and 5 s) to characterize mean position. This may have biased their findings towards short, transient force fluctuations beneath the feet, as opposed to sustained, whole-body leaning. Alternatively, the lack of gender effects in leaning may be due to a failure to control background arousal state between subjects. Factors such as time of day, or time since waking or last meal (Silver and LeSauter 2008) can influence underlying arousal state yet were not controlled across subjects and may potentially have biased the results. However, the principle focus of this study was on within-subjects changes to postural control with changes in valence and/or arousal, which should not have been influenced by background arousal state.

This study is the first to demonstrate that arousal, induced by affective pictures, can influence postural control, independent of valence. Specifically, we found that AP MPF was significantly higher in the high, compared to low, arousal conditions. This increased MPF observed across arousal conditions is similar to that reported in studies where subjects stood on an elevated surface height; a scenario that also elicits increases in physiological arousal (Brown et al. 2006; Carpenter et al. 2006; Davis et al. 2009; Huffman et al. 2009). Fear and anxiety, of which arousal is a significant component (Neiss 1988), have also been shown to increase COP frequency in individuals with phobic postural vertigo (Holmberg et al. 2003; Krafczyk et al. 1999), high trait anxiety (Wada et al. 2001), and state anxiety related to expert evaluation (Geh et al. 2011). There is a wide range of potential mechanisms through which arousal could influence postural control; yet the exact mechanisms through which arousal influences MPF are not clear. The most likely avenues for these effects are through neural projections from limbic areas to sensory modalities involved in postural control. For instance, the gain of the vestibulo-ocular response is known to be influenced by anxiety-inducing mental tasks (Collins and Guedry 1962; Yardley et al. 1995) and trait anxiety (Yardley et al. 1995); therefore, it would be reasonable to assume that descending vestibulospinal pathways involved in balance control may be similarly affected (Balaban and Jacob 2001; Balaban and Thayer 2001; Balaban 2002). Furthermore, arousal is also known to excite lower limb stretch reflexes (Bonnet et al. 1995; Both et al. 2005; Hjortskov et al. 2005; Kamibayashi et al. 2009) and improve ankle proprioceptive acuity (Matre and Knardahl 2003). It is thought that these changes reflect an arousal-driven increase in proprioceptive sensitivity. Finally, arousal has been linked to increased whole-body muscle stiffness (Fridlund et al. 1986). Such changes to proprioceptive acuity or vestibular gain could result in an increase in afferent input to spinal and/or supraspinal balance control centres. Such changes would allow the body to detect and correct for smaller amplitudes of sway and would therefore sway at a higher frequency. Alternatively, these changes might lead to increased tonic muscle activation, which would change the passive stiffness properties of the ankle, and as a result change the frequency of sway (Winter et al. 1998). Unfortunately, at this time, further research is needed to establish how these potential mechanisms are influenced by arousal and how these changes influence postural sway.

Although the exact mechanisms through which arousal (and possibly valence in the ML plane) may influence postural control are currently unknown, the observed effect of emotion on postural control has important implications for future experimental and clinical practice. From a methodological perspective, the potential confounding effects of arousal on postural control characteristics must be controlled or accounted for (a) in future studies that examine the effects of valence or other affective dimensions on postural control; (b) in longitudinal studies in which arousal states may vary, such as over the duration of an intervention, training, or rehabilitation programme (Bolmont et al. 2002; Ohno et al. 2004); and (c) when comparing between populations with different underlying levels of arousal, perhaps due to fear or anxiety, such as older adults (Arfken et al. 1994), persons with Parkinson’s disease (Bloem et al. 2001; Adkin et al. 2003), or vestibular disorders (Jacob and Furman 2001; Yardley and Redfern 2001). From a clinical point of view, there are important implications of this work as well. Firstly, older adults experience significantly more anxiety and fear and have poorer performance on clinical balance tests while performing for an expert evaluator (Geh et al. 2011). The concern is that this ‘white coat effect’, or context specific arousal, may potentially mask or mimic a balance deficit that is not normally present outside of the clinical setting. As such, we would argue that steps to reduce anxiety and arousal should be taken prior to clinical balance evaluations to maximize accuracy and external validity of such tests (Geh et al. 2011). A second clinical implication for these findings is that medications designed to suppress anxiety or arousal may have detrimental effects on postural control. Medications such as benzodiazepines, sedatives, hypnotics, and antidepressants are all associated with increased fall risk (Bloem et al. 2001; Leipzig et al. 1999; Woolcott et al. 2009). The exact mechanisms through which these medications increase fall risk are not clear; however, considering the impact of arousal on postural control in healthy subjects in this study, it would appear that the suppression of arousal through limbic inhibition would be a plausible route to investigate. In sum, we would advise that all studies of postural control take measures to monitor, and where possible control for, arousal as it may confound interpretation of experimental and/or clinical results.

We acknowledge the limitation that separating emotions into only valence and arousal is likely an oversimplification of emotion and motivation. Fontaine et al. (2007) argue that at least four dimensions (pleasantness, potency-control, arousal, and unpredictability) are required to define the emotional meanings of words. Furthermore, Harmon-Jones (2004) argues that anger and disgust, both unpleasant and arousing, activate different motivational regions of the cortex, one leading to approach and the other to withdrawal behaviour. With these points in mind, pictures in this study were chosen solely for their valence and arousal ratings and not for the emotions they are intended to invoke. We might have found more distinct contrasts between valence conditions on postural control if we used consistent emotional stimuli in each picture group. Yet, such a design would have restricted our ability to attribute our results to the broader dimensions of valence and arousal and limit us to making inferences about how specific emotional experiences influence postural control.

In conclusion, this study is the first to classify the effects of arousal, induced by affective pictures, on postural control. Furthermore, we have progressed the field by contributing data that reliably represent the full spectrum of postural sway. While our results demonstrate that arousal influences the frequency of COP displacements in the AP plane, these data do not necessarily contradict prior observations that suggest that valence can alter postural control in the ML plane if subjects are allowed to stand in narrower, less stable, stance positions. Finally, we suggest that future investigations explore the role of postural stability in emotional changes to postural control and also the mechanisms through which arousal induces changes to postural control.