Can Older Adults Resist the Positivity Effect in Neural Responding: The Impact of Verbal Framing on Event-Related Brain Potentials Elicited by Emotional Images

Abstract

Older adults have demonstrated an avoidance of negative information presumably with a goal of greater emotional satisfaction. Understanding whether avoidance of negative information is a voluntary, motivated choice, or an involuntary, automatic response will be important to differentiate, as decision-making often involves emotional factors. With the use of an emotional framing event-related potential (ERP) paradigm, the present study investigated whether older adults could alter neural responses to negative stimuli through verbal reframing of evaluative response options. The late-positive potential (LPP) response of 50 older adults and 50 younger adults was recorded while participants categorized emotional images in one of two framing conditions: positive (“more or less positive”) or negative (“more or less negative”). It was hypothesized that older adults would be able to overcome a presumed tendency to down-regulate neural responding to negative stimuli in the negative framing condition thus leading to larger LPP wave amplitudes to negative images. A similar effect was predicted for younger adults but for positively valenced images such that LPP responses would be increased in the positive framing condition compared to the negative framing condition. Overall, younger adults' LPP wave amplitudes were modulated by framing condition, including a reduction in the negativity bias in the positive frame. Older adults' neural responses were not significantly modulated even though task-related behavior supported the notion that older adults were able to successfully adopt the negative framing condition.

Across many sub-domains of psychological research, older adults have been shown to exhibit a shift in emotional prioritization away from negatively valenced mood and information and towards positively valenced mood and information (Mather & Carstensen, 2005; Murphy & Isaacowitz, 2008; Reed & Carstensen, 2012). For example, older adults appear to avoid negative states, engage in more passive styles of problem solving which are related to less negative mood, and attempt to keep a positive mood when faced with negative information (Blanchard-Fields, Stein & Watson, 2004; Isaacowitz, Toner, Goren, & Wilson, 2008; Lockenhoff & Carstensen, 2004; Mather, Canli, English, Whitefield, Wais, Ochsner, et al., 2004). This shift in emotional bias has also been documented in the domains of memory (Charles, Mather, & Carstensen, 2003; Leigland, Schulz, & Janowsky, 2004; Mikels, Larkin, Reuter-Lorenz, & Carstensen, 2005), attention (Isaacowitz, Toner, Goren, & Wilson, 2008; Mather & Carstensen, 2003; Rösler et al., 2005) and even measures of regional cerebral blood flow (Brassen, Gamer, & Büchel, 2011; Leclerc & Kensinger, 2011; Mather et al., 2004; Williams, Brown, Palmer, Liddell, Kemp, Olivieri, et al. 2006).

Brain electrical activity evoked by emotional stimuli can be measured non-invasively with event-related brain potentials (ERPs), and this approach has also been applied to the study of age-related changes in the prioritization of emotional information (Kisley, Wood, & Burrows, 2007; Langeslag & van Strien, 2009). ERPs are time sensitive, real-time measurements of neuron population activity, and can be used to capture neuronal activity before or in the absence of overt behavior (Stern, Ray, & Quigley, 2001; Kramer, Fabiani, & Colcombe, 2006, Luck & Girelli, 1998). The late positive potential (LPP) waveform is one specific ERP component, typically evoked by emotional visual images, that is largest over the parietal scalp area (Cacioppo & Berntson, 1994). Maximal LPP wave activity occurs between 400 msec and 900 msec post-stimulus presentation, and in response to motivationally relevant images (Schupp, Cuthbert, Bradley, Cacioppo, Ito, & Lang, 2000; Weinberg & Hajcak, 2010). In a study by Ito, Larsen, Smith, and Cacioppo (1998), younger adult participants viewed positive and negative target images embedded in a neutral context while the LPP waveform was recorded. The largest LPP amplitude occurred when younger adults viewed the negative images compared to the positive images. Ito et al. (1998) described this as a physiological reflection of the negativity bias. Wood and Kisley (2006) repeated this study but with both older and younger adults. They found similar results for younger adults; however, older adults did not show a bias to either negative or positive stimuli. A follow-up study revealed that this effect resulted from age-related decrease in responding to negative images with relative stability in LPP amplitudes in response to positive images (Kisley et al., 2007). Langeslag and van Strien (2009) showed that this electrophysiological effect was related to age-related changes in memory for emotional stimuli.

Such “positivity effects” seen in older adults have often been explained through Socioemotional Selectivity Theory (SST), which posits that an individual's goals and therefore motivations are influenced by the perceived amount of time that is left before death (Carstensen, 1992). Healthy younger adults view time as more expansive and thus are more motivated to seek out new information to further development (Carstensen & Mikels, 2005). With age, one becomes less concerned with novel information and goals shift such that there is more concern with maintaining satisfying social relationships and staying emotionally regulated (Carstensen & Mikels, 2005; Carstensen, Isaacowitz, & Charles, 1999; & Carstensen, 2005). Importantly, positivity effects in responding are conceptualized within this model to result from voluntary efforts by older adults aimed at improving overall feelings of well-being (Carstensen, Issacowitz, & Charles, 1999) and there is evidence to support this idea (Knight et al., 2007; Isaacowitz, Toner, & Neupert, 2009). In contrast, some have argued that negative stimuli require a higher degree of effortful processing (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001) so it is possible that positivity effects in older adults are due to cognitive decline and therefore an involuntary developmental shift (Calder, Keane, Manly, Sprengelmeyer, Scott, Nimmo-Smith et al. 2003; Ruffman, Henry, Livingstone, & Phillips, 2008). Although comparatively little direct evidence has been garnered for this latter perspective, positivity effects in attention, memory and brain function have also not been universally detected in studies of older adults (Kensinger, Brierly, Medford, Growdon, & Corkin, 2002; Mickley Steinmetz, Muscatell, & Kensinger, 2010) raising the possibility that their expression may depend on study context. Put another way, positivity effects may be flexible in older adults depending on their immediate motivational goals (Reed & Carstensen, 2012). For example, under experimental conditions when older adults are instructed to reduce prioritization of emotional regulation goals, the positivity effect does not appear as strong (Löckenhoff & Carstensen, 2007). The present study was designed to more directly examine the potential flexibility of emotional biases in the neural responses of older adults. Can the positivity effect in LPP responses to emotional images be overcome by altering the task instructions? To address this question we manipulated a variable previously shown to affect such responses in younger adults: the evaluative framework within which the stimuli are presented (Kisley, Campbell, Larson, Naftz, Regnier, & Davalos, 2011).

It has been shown previously that LPP amplitude to emotional stimuli can be modulated through voluntary efforts, at least in younger adults. Attentional modulation, including overall attention (Thiruchselvam, Blechert, Sheppes, Rydstrom, & Gross, 2011) as well as attention directed specifically to a stimulus's emotionality (Ferrari, Codispoti, Cardinale, & Bradley, 2008), can increase LPP amplitude. Additionally, the emotional regulation strategy of cognitive reappraisal has also been investigated with ERPs (Hajcak & MacNamara, 2012; Moser, Hajcak, Bukay, & Simmons, 2006). When younger adults were asked to reinterpret the meaning of pleasant, neutral, or unpleasant pictures, Hajcak and Niewenhuis (2006) found that LPP wave amplitudes were reduced in response to pleasant and unpleasant pictures. The reappraisal instructions given by Hajcak and Niewenhuis were meant to change the way a participant related to that image. For example, before a negative image was viewed, the participant was given the instruction to reinterpret that image. By reappraising a negative picture, the participant was able to modulate a negative response. In another study that examined how reappraisal could modulate LPP waves, Foti and Hajcak (2008) gave two types of negative descriptions of the images to participants. One description was more negative and the other description was less negative. After participants listened to these descriptions, negative images were presented. The LPP waves associated with the more negative descriptions were greater in amplitude than the waves elicited by the less negative descriptions. The reappraisal frame of the more negative description was related to an increased LPP wave amplitude (Foti & Hajcak, 2008). A similar finding occurred in a recent study that employed a modified framing approach (Kisley, Campbell, Larson, Naftz, Regnier, & Davalos, 2011). In a between-subjects design, a negative frame group was asked to evaluate whether an image was “negative” or “not negative.” In the positive frame group, participants viewed the same images but evaluated whether each picture was “positive” or “not positive.” Kisley et al. discovered that the negative frame group exhibited significantly greater LPP responses for negative images than the positive frame group. Thus, this form of verbal framing can influence brain responses to emotional stimuli. However, this has not been tested in older adults.

By using a framing paradigm for the present investigation, older adults were asked to exchange their presumed positive responding mode for a more negative one. If older adults are voluntarily favoring positive stimuli over negative stimuli, we expected that they would be able to overcome this bias in neural responding (in this case the LPP waveform). Several specific predictions were made. A main effect of age was predicted (Wood & Kisley, 2006), such that younger adults overall were expected to garner larger LPP wave amplitudes than older adults. A main effect of valence was predicted (Schupp et al., 2000l; Kisley et al., 2007; Weinberg & Hajcak, 2010), such that emotional images were expected to elicit larger LPP wave amplitudes than neutral images overall. A frame × valence interaction was predicted based on previous findings from Kisley et al. (2011). Specifically, LPP wave amplitudes for negative images in the negative framing condition were expected to be larger than positive or neutral images and the opposite was expected for the positive framing condition. If the two age groups respond to the framing paradigm differently, then an age × frame × valence interaction would be expected as well. By contrast, following the theory of SST, older adults should be able to shift their responding when asked to adopt a negative evaluative frame, thus exhibiting a pattern of modulation that is similar to the younger adults.

Method

Participants

Younger adults were recruited from the University of Colorado at Colorado Springs (UCCS) and received course credit for their participation. Older adults were recruited from the UCCS Gerontology Center Registry and were monetarily compensated for their participation.

Overall, a total of 56 younger adults were recruited for the study. Six participants were excluded due to insufficient number of trials available for computing ERP average waveforms (n = 3) and electrophysiological recording problems (n = 3; details below). Statistical analyses were performed on the remaining 50 younger adult participants whose ages ranged from 18 to 27 years old (M = 21.08, SD = 2.60; 35 females; 78% White, 10% Latino, 6 % African-American, and 6% Asian or Pacific Islander). All included younger adult participants tested 20/40 or better with corrected vision on the Snellen visual acuity chart and were screened by self-report for current use of psychoactive drugs including antidepressants.

A total of 60 older adults were recruited for the study. Ten of these participants were excluded due to insufficient number of trials available for computing ERP average waveforms (n = 3), electrophysiological recording problems (n = 4), vision difficulty (n = 1), and scores on the Dementia Rating Scale (DRS) that fell below the cutoff (n = 2). Statistical analyses were performed on the remaining 50 older adults whose ages ranged from 56 to 83 years old (M = 66.20, SD = 6.45; 31 females; 96% White and 4% Latino). All included older adult participants tested at least 20/40 or better with corrected vision on the Snellen visual acuity chart and were screened by self-report for current use of psychoactive drugs including antidepressants.

All participants were alternately counterbalanced into either the positive framing condition or the negative framing condition: younger positive frame mean age of 20.92 (SD = 2.38) years; younger negative frame mean age of 21.24 (SD = 2.85) years; older positive frame mean age of 66.80 (SD = 5.81) years; older negative frame mean age of 65.60 (SD = 7.10) years.

Materials

Self Assessment Manikin Instrument

The Self Assessment Manikin instrument (SAM: Lang, Bradley, & Cuthbert, 2005) was used to collect subjective ratings of the images utilized to elicit the ERPs. To use the SAMs, participants ranked the target images that they viewed on two different picture-based scales after the ERP recording was completed. Participants first indicated the valence of each image (from 1, most negative to 9, most positive, with 5 being neutral) followed by the arousal level of each image (from 1, most calm, to 9, most arousing). As described below, this measure was completed after the ERP paradigm was completed.

ERP Paradigm Materials

Participants viewed emotional images presented on a 17-in. liquid-crystal display color computer monitor approximately 2.5 ft (0.76 m) from the participant. E-Prime (Psychological Software Tools, Inc., Pittsburgh, PA) was used to present the images and the framing conditions as well as record behavioral responses (categorizations and response times) as participants viewed and categorized the images. Positive, negative, and neutral images were selected based on their published subjective ratings from the International Affective Picture System (IAPS; Lang, Bradley, & Cuthbert, 2005). ERP responses were recorded for the following: negative images 2703 (sad children), 9902 (car accident), 7380 (roach on pizza); positive images 1710 (puppies), 2216 (children at a swimming pool), 7502 (Disney World castle); neutral images: 2397 (men on a subway), 2840 (a child playing chess), 7207(beads). Additionally, there were 36 neutral images (ordinary household objects, buildings, etc.) used to set a neutral context for the odd-ball paradigm¹. The ERP responses to these images were not analyzed. The electrophysiological signals were recorded on a Neuroscan NuAmps amplifier system using Scan 4.2 installed on a laptop computer. Scan 4.2 software was used for electrophysiological signal acquisition and analysis. Participants used button presses on a computer mouse to categorize the images. All images used for ERP analysis were selected for approximately balanced luminance and image complexity.

The means and standard deviations of the subjective valence and arousal ratings for both younger and older adults across positive and negative framing conditions can be seen in Table 1. There was no significant difference between how the four groups (older positive frame, older negative frame, younger positive frame, younger negative frame) rated the emotional images (negative images: valence rating F(3, 96) = .60, p = .61, and arousal level F(3, 96) = 1.66, p = .18, positive images: valence rating: F(3, 96) = .29, p = .83, arousal level: F(3, 96) = 1.45, p = .23, neutral images: valence rating: F(3, 96) = 1.60, p = .20, arousal level: F(3, 96) = 1.62, p = .19). To ensure that positive and negative images were experienced as equally valenced, mean absolute-value differences from neutral were computed and compared across all participants. The mean absolute-value difference between the negative images' valence ratings and theoretical neutral (i.e., 5) was 3.62 (i.e., |1.36 - 5|), which was not significantly different from the distance for the positive images' mean valence ratings and neutral which was 3.53 (i.e., |8.50 - 5|), [F(1, 97) = 2.06), p = .15]. Balance in arousal ratings between positive and negative images was also assessed. The overall negative image arousal rating (M = 6.04) was not significantly different from the overall positive image arousal rating [M = 6.58; t(99) = 1.51), p = .14].

Table 1. Means and standard deviations for subjective valence and arousal ratings (from the SAMS) of positive, negative, and neutral images for younger and older adults in both the positive and negative framing conditions.

		Positive Framing Condition		Negative Framing Condition
		Valence Ratings M(SD)	Arousal Ratings M(SD)	Valence Ratings M(SD)	Arousal Ratings M(SD)
Younger Adults	Neutral Images	5.47(1.12)	2.73(1.21)	5.41(1.06)	2.77(1.37)
	Positive Images	8.35(.88)	7.12(1.54)	8.59(.63)	6.19(1.88)
	Negative Images	1.19(.41)	5.68(2.60)	1.48(1.05)	6.00(2.11)
Older Adults	Neutral Images	5.99(1.55)	3.53(1.52)	6.05(1.52)	3.09(1.59)
	Positive Images	8.47(1.22)	6.80(1.97)	8.47(.82)	6.20(2.23)
	Negative Images	1.39(.91)	5.52(2.85)	1.37(.64)	6.96(2.40)

Dementia Rating Scale

All older adult participants were screened using the Dementia Rating Scale (DRS: Mattis, 1988). The DRS is an individually administered assessment designed to measure the level of cognitive functioning for older adults. A cutoff score of 6, which is one standard deviation below the mean, was used in this study because it has been found that dementia-free individuals who fall more than one standard deviation below the mean on a test of cognitive ability are more likely to later be diagnosed with dementia (Buse, Hensel, Guhne, Angermeyer, & Riedel-Heller, 2006).

Procedure

Upon agreement of participation over the telephone, older adults were given verbal instructions for getting to the electrophysiology lab at the university. Younger adults were asked to come to the lab during their scheduled appointment. Upon arrival participants were informed of their rights, the procedures for the study, and they were asked to sign an informed consent document. Older adults were screened for cognitive deficits using the DRS. Regardless of scores achieved on the DRS, all older adults continued with study procedures. Once both the younger and older adults were enrolled into the research study, they were screened for vision difficulties. Participants were alternately counterbalanced by assignment to the positive framing group or the negative framing group.

Participants were prepared for electrophysiological assessment following the locations of the International 10-20 system. Silver/silver-chloride electrodes were attached to Fz, Cz, Pz, right mastoid, and left mastoid. Electrodes were attached above and lateral to the participant's left eye to monitor eye movements. All electrodes were referenced to the left mastoid, and a ground electrode was attached to the forehead. Electroencephalographic activity was band-pass filtered (0.1-100 Hz) and sampled at 1000 Hz. Impedance values were monitored to ensure a maximum impedance value of 10kΩ. Brain-wave signals were collected and digitized on a multi-channel recording system (Nu-Amps amplifier, Neuroscan Inc.).

Before the paradigm began, participants were provided instructions and shown example images. A script was read to each participant which described that a picture would be presented followed by a screen with words “Positive” (or “Negative,” depending on group assignment) at the top and the words “more” and “less” on the bottom. Participants were instructed to respond according to their general impressions to each image, regardless of the other images encountered. Participants completed five practice trials which helped orient them to the task and the type of images they would be encountering. After the practice session, participants were given an opportunity to ask questions concerning task instructions.

During the paradigm, each participant viewed each image for 1 sec and then the response screen appeared. In the negative frame, the words “Negative” (top half of screen), “more” or “less” (bottom half of screen, corresponding to left or right response buttons) were shown. In the positive frame, the words “Positive,” “more” or “less” appeared on the screen. If participants did not respond within 5 seconds, the trial was classified as a “non-response” and the paradigm continued to the next block of images. The time that it took participants to categorize images by pressing the mouse button was captured to the nearest millisecond and analyzed, as were the response selections.

Regardless of framing group, each participant viewed the same series of images. Images were shown in 90 blocks of five pictures each (450 total image presentations). There was a brief pause after each block, which was terminated by a button press. Overall, ERP recordings were measured for the 90 target image presentations (30 neutral, 30 positive, and 30 negative), corresponding to 10 presentations for each of the target images described above under materials. The remaining 360 image presentations consisted of neutral images selected from the IAPS (36 images, each presented 10 times). This pattern of presentation was intended to set a “neutral context” within which the target images were presented (Ito et al., 1998; Kisley et al., 2007). Target images were presented pseudo-randomly in either the third, fourth, or fifth position in each block. The SAMs ratings for each target image were collected after the paradigm was completed.

Analyses

ERP analysis

The ERP analysis employed here was identical to that of Wood and Kisley (2006) and Kisley et al. (2007). For each of the 90 target image presentations, stimulus-locked ERP intervals (epochs) were computed from the electroencephalographic recording from100 ms before to 900 ms after image presentation. The epochs were re-referenced to an electrode computed as the average of left and right mastoids. If any channel exceeded ±100 μV during an epoch, the trial was assumed to be corrupted due to a movement artifact and thus it was excluded from further analysis. After corrupted trials were removed, the remaining trials were sorted by valence (positive, negative, and neutral) and average waveforms were computed and low-pass filtered at 9 Hz. Participants with fewer than ten single trials (of 30 available) for any valence category were excluded from further analyses. The mean number of trials used to compute average ERP waveforms did not differ between the four groups [F(3,96) = 1.46, p = .23].

LPP amplitude is generally largest over the Pz recording site (Ito et al., 1998). Therefore LPP amplitude and latency for each participant's positive, negative, and neutral waveforms were taken from the largest peak voltage between 400 and 900 ms post-image onset (Coles, Gratton, & Fabiani, 1990; Kisley et al., 2007).

Statistical analyses

All statistical analyses were computed using PASW Statistical Software version 18.0 (SPSS, 2009). Several 2×2×3 mixed design analysis of variances (ANOVAs) were performed. Post-hoc analyses were performed using Fisher's least significant difference (LSD). Age (younger adult and older adult) and framing condition (positive frame and negative frame) served as the between-groups independent variables while image valence (positive, negative, and neutral) served as the within-group independent variable. LPP peak wave amplitude was the primary dependent variable of interest here; other dependent variables of interest were behavioral response time in milliseconds, and response consistency (percent of image categorizations consistent with expectation based on published norms). The primary purpose of analyzing response consistency was to ensure that participants were complying with the task. Response consistency was evaluated based on whether a participants' categorization of an image was consistent with published norms for that image. For example, if a participant categorized a positive image as “more” positive in the positive framing condition or “less” negative in the negative framing condition, these responses were considered to be consistent with expectation. The response consistency score was based on the number of times out of 30 that participants categorized either a positive, negative, or neutral image consistent with expectation. We also examined the response consistency of neutral images in order to investigate potential behavioral differences between the framing conditions. Expectation for neutral images was that they would be rated as “less positive” and “less negative.”

While recording behavioral responses (i.e., response time and response consistency) a total of n = 8 younger adult participant's data were lost due to a systematic data collection error; specifically the E-prime data files were inadvertently overwritten. Therefore, the analyses for behavioral response time and response consistency was based on a sub-sample of the original 50 younger adults reported above: n = 42 (32 female) younger adults with an average age of M(SD) = 21.17(2.60). However, the ERP data and the subjective SAMS ratings on image valence and arousal from these eight participants were still available and included in the final analysis for these variables.²

Results

Behavioral Analyses

Both age and image valence affected behavioral response times. The mean response times for younger and older adults separated by framing condition are presented in Table 2. A 2×2×3 mixed design ANOVA was performed to assess how the response times of older and younger adults differed across positive and negative framing conditions while looking at positive, negative, and neutral images. Mauchly's test indicated that the assumption of sphericity had been violated; Mauchly's W = .786, χ² (2) = 20.898, p < .001, therefore degrees of freedom were corrected using the Huynh-Feldt estimates of spericity (ε = .87). As expected there was a main effect of age, F(1, 88) = 22.68, p < .001, η² = .21. When both framing conditions were combined and all image types collapsed, younger adults were significantly faster at responding (M = 594.10 msec) than older adults (M = 835.48 msec, p < .001). There was also a significant main effect of image type, F(1.73, 152.48) = 12.12, p < .001, η² = .12. Post-hoc comparisons revealed that when both age groups and framing conditions were combined, response times for neutral images (M = 777.37 msec) were significantly slower than the response times for both positive (M = 688.93 msec, p < .001) and negative (M = 678.07 msec, p < .001) images.

Table 2. Means and standard deviations for response times and consistency for younger and older adults in both the positive and negative framing conditions.

		Positive Framing Condition		Negative Framing Condition
		Response Times (ms) M(SD)	Response Consistency (%) M(SD)	Response Times (ms) M(SD)	Response Consistency (%) M(SD)
Younger Adults	Neutral Images	653.87 (283.33)	42.40 (27.12)	652.25 (209.62)	87.68 (17.71)
	Positive Images	551.33 (218.60)	94.27 (11.85)	605.10 (204.77)	99.71 (1.39)
	Negative Images	559.45 (188.95)	99.47 (1.58)	542.61 (218.60)	98.82 (2.21)
Older Adults	Neutral Images	938.52 (390.09)	23.23 (23.47)	864.84 (271.91)	88.95 (23.86)
	Positive Images	822.91 (309.56)	99.87 (0.67)	776.40 (260.77)	97.72 (8.39)
	Negative Images	823.07 (302.92)	94.53 (11.09)	787.14 (310.74)	91.40 (20.95)

Participants in the negative framing condition were significantly more consistent in their categorizations of valenced images than participants in the positive framing condition. The means and standard deviations of the response consistency of younger and older adults in the positive and negative framing conditions are presented in Table 2. A 2×2×3 mixed design ANOVA was performed to assess the response consistency. The dependent variable (percent consistent) was extremely skewed, so a nonlinear transformation (base 10 log of percent consistent) was performed. Mauchly's test indicated that the assumption of sphericity had been violated; Mauchly's W = .356, χ² (2) = 77.412, p < .001, therefore degrees of freedom were corrected using the Greenhouse-Geisser estimates of spericity (ε = .61). There was a significant main effect of framing condition, F(1, 76) = 24.88, p < .001, η² = .25. Across age groups and all image types, participants in the negative framing condition rated images more consistently (M = 94.05%) than participants in the positive framing condition (M = 75.62%, p < .001). There was also a main effect of image type, F(2.71, 92.47) = 57.56, p < .001, η² = .43. Post-hoc comparisons revealed that across both age groups and both framing conditions, response consistency for neutral images was significantly lower (M = 60.56%) than for positive (M = 97.89%, p < .001) and negative (M = 96.06%, p < .001) images. Finally, there was a significant image by frame interaction, F(1.22, 92.47) = 31.08, p < .001, η² = .29. LSD-adjusted post-hoc comparisons revealed participants in the negative frame categorized neutral images more consistently (M = 88.31%) than participants in the positive frame (M = 32.81%, p < .001).

Electrophysiological Analyses

The LPP peak wave amplitudes differed between age groups and framing conditions. A 2×2×3 mixed design ANOVA was performed to assess how the LPP wave amplitudes of older and younger adults differed across positive and negative framing conditions when looking at positive, negative, and neutral images. Table 3 shows the means and standard deviations for LPP peak wave amplitudes. Figure 1 illustrates the LPP grand averaged waveforms for both younger and older adults in each framing group while viewing neutral, positive and negative images. Mauchly's sphericity test was not significant: Mauchly's W = .969, χ² (2) = 3.033, p = .219. Therefore no correction was made to the degrees of freedom used to evaluate the significance of the F ratio. As expected, there was a significant main effect of age, F(1, 96) = 4.36, p = .04, η² = .04. Across both framing conditions and for all image types, the younger adults' LPP wave amplitudes (M = 8.16 μV) were significantly larger than the older adults' LPP wave amplitudes (M = 6.60 μV). There was also a significant main effect of image valence, F(2, 192) = 27.63, p < .001, η² = .22. Post-hoc comparisons revealed that for younger and older adults collapsed across framing conditions, neutral images produced significantly smaller (M = 5.61 μV) LPP wave amplitudes than both positive (M = 8.40 μV) and negative (M = 8.13 μV) images. A significant two-way interaction emerged between age and image valence, F(2, 192) = 18.40, p < .001, η² = .16. After completing a post-hoc comparison to compare image valences, the results demonstrated that younger adult's LPP wave amplitudes for neutral images (M = 5.17 μV) were significantly smaller (p <.001) than those for positive (M = 9.12 μV) or negative (M = 10.19 μV) images collapsed across framing conditions. The LSD-adjusted post-hoc comparison for older adults revealed that across framing conditions, the LPP wave amplitude elicited by positive images (M = 7.67 μV) was significantly larger than neutral (M = 6.06 μV, p < .01) and negative (M = 6.07 μV, p < .01) images; a positivity effect in the older adults.

Table 3. Means and standard deviations for LPP peak wave amplitude and latency to positive, negative, and neutral images for younger and older adults in both the positive and negative framing conditions.

		Positive Framing Condition		Negative Framing Condition
		LPP Peak Wave Amplitude (μV) M(SD)	LPP Peak Wave Latencies (ms) M(SD)	LPP Peak Wave Amplitude (μV) M(SD)	LPP Peak Wave Latencies (ms) M(SD)
Younger Adults	Neutral Images	5.83 (3.71)	562.96 (67.02)	4.50(4.62)	552.96 (67.61)
	Positive Images	10.77(5.34)	514.80 (55.49)	7.47(5.57)	519.08 (75.04)
	Negative Images	8.92(5.91)	544.84 (41.34)	11.46(6.34)	532.68 (43.70)
Older Adults	Neutral Images	6.91(3.13)	582.60 (75.78)	5.20(3.36)	546.68 (107.60)
	Positive Images	8.24(3.76)	554.28 (69.05)	7.10(3.06)	534.80 (92.20)
	Negative Images	6.52(2.62)	593.52 (77.62)	5.63(3.76)	577.20 (90.11)

Figure 1.

Across-subjects grand averaged ERP waveforms for negative, neutral, and positive images separated for younger and older adults in negative and positive framing conditions. These waveforms are presented for illustration purposes, but not analyzed. The LPP is the prominent positive waveform peaking between about 500 and 600 ms.

A significant two-way interaction was also discovered between image valence and framing condition, F(2, 192) = 7.43, p < .001, η² = .07. An LSD-adjusted post-hoc comparison revealed that in the positive frame for younger and older adults combined, the LPP wave amplitudes produced by positive images (M = 9.51 μV) were significantly larger than those for negative (M = 7.72 μV, p < .01) and neutral images (M = 6.37 μV, p < .001). The LPP wave amplitudes for negative images were significantly larger than those for neutral images (p < .05). In the negative frame, LPP wave amplitude for negative images (M = 8.54 μV) was significantly larger than for positive (M = 7.29 μV, p < .05) and neutral (M = 4.85 μV, p < .001) images. Furthermore within the negative framing condition, the LPP wave amplitudes for positive images were significantly larger than those for neutral images (p < .001). In sum, the largest LPP responses occurred in the congruent framing condition.

Finally, a significant three-way interaction was discovered between age, framing condition, and image valence, F(2, 192) = 5.70, p < .01, η² = .06. Post-hoc comparisons revealed that younger adults in the positive frame had significantly larger LPP wave amplitudes when presented with positive images [M = 10.77 μV, 95% CI(8.68, 12.75)] than when presented with negative [M = 8.92 μV, 95% CI(7.21, 11.60), p < .05] or neutral [M = 5.83 μV, 95% CI(4.21, 7.72), p < .001]. Conversely, the LPP wave amplitudes of younger adults in the negative frame were significantly larger for negative images [M = 11.46 μV, 95% CI(8.98, 13.37)] as opposed to both positive [M = 7.47 μV, 95% CI(5.42, 9.49), p < .001] and neutral [M = 4.50 μV, 95% CI(2.51, 6.02), p < .001] images. In other words the largest response corresponded to the image valence that was congruent with the frame within the younger adult group. For older adults in the positive framing condition, positive images [M = 8.24 μV, 95% CI(6.47, 10.02)] elicited significantly larger LPP wave amplitudes than negative images [M = 6.52 μV, 95% CI (4.60, 8.43), p < .05]. The primary effect driving the 3-way interaction: older adults did not show a reversal of this pattern in the negative frame. Instead, the only finding in this condition was that older adults exhibited significantly larger LPP wave amplitudes while looking at positive images [M = 7.10 μV, 95% CI(5.32, 8.87)] as compared to neutral images [M = 5.20 μV, 95% CI(3.68, 6.73), p < .05]. But overall their pattern of responding was similar to the negative frame (see Figure 2).

Figure 2.

Mean LPP peak wave amplitudes elicited by neutral, positive and negative images, separately for younger and older adults in the positive and negative framing conditions. Note the interaction effect whereby maximal response amplitude is sensitive to framing condition in the younger adults (i.e., negative responses are largest in the negative frame, positive responses in the positive frame) but not the older adults (positive responses are largest in both frames).

Discussion

The current study explored the effect of differential framing of evaluative categorization language on the neural responses elicited by emotional images in younger and older adults. The results provide further evidence that such verbal framing can be effective at modulating the LPP wave amplitudes of younger adults (Kisley et al., 2011). In fact, this group showed a reversal of their apparent negativity bias in neural responding (Ito et al., 1998), as responses to positive images were significantly larger than to negative images in the positive frame. In the current paradigm older adults were not effective at differentially modulating their LPP wave amplitudes by framing condition like their younger counterparts. In fact, older adults maintained a positivity effect across framing conditions. Due to existing theory and evidence suggesting that older adults typically prioritize positive stimuli through voluntary efforts (reviewed by Reed & Carstensen, 2012), it was expected that they would be able to overcome this response mode in the negative framing condition here, thus leading to larger LPP wave amplitudes to negative images like their younger counterparts. Potential explanations for the apparent lack of response modulation for the older adults group are considered below.

Beyond the novel finding described above, the present results were largely consistent with previous ERP investigations of neural responses to emotional images including age-related changes. For example, younger adults demonstrated significantly larger LPP waveforms overall than older adults. This pattern of age-related neural dampening has been observed in previous studies of both the LPP in response to emotional images (Wood & Kisley, 2006), as well as related ERP components such as the P300 in response to more rudimentary stimuli such as simple visual shapes and auditory sounds (reviewed by Kok, 2000). The observed increase in LPP wave amplitude in response to emotional images compared to neutral images was also expected. Emotional stimuli grab attention and elicit larger LPP wave amplitudes than neutral images (Cuthbert et al., 2000; Hajcak & MacNamara, 2010; Ito et al., 1998; Schupp et al., 2000).

The significant interaction between image valence and framing condition confirmed that, with the use of different verbal framing of response options, LPP wave amplitude can be modulated (Kisley et al., 2011). Overall, positive images in the positive framing condition elicited larger LPP peak amplitudes, and negative images in the negative framing condition elicited larger LPP peak amplitudes. This result is important for a number of reasons. First, the emotional frame used in this study was not an explicit framing task as in previous studies of emotion regulation (e.g., Foti & Hajcak, 2008; MacNamara, Ochsner, & Hajcak, 2011; Moser, Hajcak, Bukay, & Simons, 2006), but rather an implicit reframing of how an image is categorized. Participants in this study were not explicitly instructed to try to actively enhance their negative affect in the negative framing condition or their positive affect in the positive framing condition. They were simply provided with an anchor (i.e., “Negative” or “Positive”) and asked to categorize images based on that word. It was shown here that doing so can alter the balance in brain responses between the two emotional extremes. Previous studies using implicit instructions to examine LPP amplitude have shown larger wave amplitudes when participants were asked to categorize images on an affective dimension versus a non-affective dimension (Hajcak, Moser & Simons, 2006). Although there was an effect for the implicit affective decision in that study, there was no differentiation between the LPP wave amplitude for positive or negative images. The presently demonstrated differential modulation of images based on their valence is also noteworthy because of the significant increase in LPP wave amplitude specifically for positive images. Although many studies have focused on efforts to modulate LPP magnitude in response to negative stimuli (Foti & Hajcak, 2008; Dunning & Hajcak, 2009; recently reviewed by Hajcak & MacNamara, 2010), few have been aimed at modulating the response to positive stimuli separately from negative stimuli. One previous framing investigation failed to uncover significant modulation of LPP responses to positive images in younger adults (Kisley et al., 2011). However, here we employed a subtle, yet potentially important difference in framing approach: participants in the positive frame rated images as “more” or “less” positive. Previously they were asked to rate images by selecting “yes” or “no” in response to the prompt “Positive?” (Kisley et al., 2011). Although it's unclear exactly why the former led to significant modulation and the latter did not, we believe that the “more” judgment better highlights the discrepancy between the positive image and the ongoing neutral context of the paradigm. In support of this idea, it has recently been shown that the relative extent to which an emotional image is a target within a task – that is, its rarity of presentation and corresponding behavioral response selection within a paradigm – tends to increase LPP amplitude (Ferrari et al., 2008; Weinberg, Hilgard, Bartholow, & Hajcak, 2012). Indeed, this general effect may represent the mechanism by which verbal framing leads to modulation of LPP amplitude in younger adults: priming individuals with a particular emotional frame (“positive” or “negative”) may lead to heightened and selective anticipation of the corresponding emotional images and associated rare affirmative response selection, thus leading to greater neural responses to oddball emotional image when it occurs. However, this idea remains to be tested directly in the context of verbal framing manipulations.

The interaction found between age and image valence extends previously-demonstrated differences in the balance of neural responding to positive and negative images between older and younger adults. Here, older adults exhibited a stronger neural response to positive images than to negative images (i.e., a “positivity effect”) when compared to younger adults. This result is notable, as there has not been strong evidence demonstrating a positivity effect in the LPP waveform in any age group. Rather, previous studies have shown “negativity biases” present in the youngest adults (around 21 years of age; Ito et al., 1998) shift during adult development into more balanced responding between positive and negative images (Wood and Kisley, 2006). This has been interpreted as a relative lack of emotional bias (Kisley et al., 2007). Differing image content across studies might explain the discrepancy between the present and previous investigations. For example, previous studies utilized images of appetizing foods for positive stimuli, and images of dead animals for negative stimuli. Here, a more diverse array of image themes was employed. It has been demonstrated recently that image theme can modulate LPP amplitude (Weinberg and Hajcak, 2010). Also, the subjective ratings of positive and negative images, both for valence and arousal, were statistically balanced in the present study, whereas the negative images have generally been more valenced and arousing than positive images in previous studies. This could have led to masking of any positivity effects inherent in the older adult samples from those studies. Regardless, the present study is consistent with past research in that the neural responses exhibited a pattern of age-related shifting from negative, towards positive prioritization.

Finally the significant three-way interaction between age, image type, and framing condition demonstrates that younger and older adults responded to the framing conditions differently when viewing positive, negative, and neutral images. While younger adults' LPP wave amplitudes were highly sensitive to their framing condition, the older adults' LPP wave amplitudes were not. When younger adults viewed positive images in the positive framing condition, the resulting LPP wave amplitudes were significantly larger than when their counterparts in negative framing condition viewed the same positive images. In addition, the LPP wave amplitudes produced by younger adults in the negative framing condition while viewing negative images were significantly larger than those produced while they viewed positive or neutral images. By contrast, older adults did not exhibit significant modulation of neural response to valenced images based on the emotional frame.

A relatively trivial explanation for the lack of response modulation across framing conditions for older adults would be that this group simply did not successfully adopt the assigned response frameworks. However, the pattern of behavioral responses provides evidence against this interpretation. Like the younger adults, the older adults exhibited a pattern of response consistency that matched the assigned frame: that is, their categorizations were greater than 90% consistent with expectation for both positive and negative images across frames. Also telling is the response consistency for neutral images. Similar to their younger counterparts, the older adults exhibited relatively low consistency in the positive frame - specifically, they rated neutral images as “less positive” about 23% of the time - but substantially higher consistency in the negative frame, rating neutral images as “less negative” nearly 90% of the time. Although this observation supports the idea that the older adults were performing the task like the younger adults, it does not address the potential cause of this rating asymmetry. One possible explanation is the positivity offset: when an individual is under low levels of threat, exploratory behavior increases and neutral stimuli are often evaluated as slightly more positive (Ito & Cacioppo, 2005; Norris, Gollan, Berntson, & Cacioppo, 2010). Positively valenced evaluation of neutral stimuli motivates individuals to approach and investigate novel information. Thus, it is possible that participants in the positive framing condition were more contemplative about the neutral images and more likely to report them as being “more positive,” against our a priori expectation that they would be rated as “less positive.” In support of this idea, response times were longest for the neutral images in the present study. There was also an important difference in behavioral performance between age groups: older adults demonstrated slower response times than the younger adults overall. This slower reaction time is consistent with the pattern of normal cognitive aging. As individuals age, processing speed begins to decline along with several other cognitive abilities including working memory (Schaie & Zanjani, 2006; Salthouse, 1996). Interpreted as such, and because there was not an apparent differential age-related slowing across framing conditions or image valences (that is, there were no significant interaction effects involving age-group for any behavioral variable), this finding does not appear to go against the contention that older adults had successfully adopted the verbal framing task here.

In conclusion, the results from this study demonstrate an apparent robustness of the positivity effect in neural responding in older adults across verbal framing conditions. It is possible that, contrary to SST, older adults' prioritization of positive information (and/or avoidance of negative information) could be a relatively automatic process, at least at the level of LPP responses to emotional images. If so, substantial modulation of brain responses across framing conditions would not be expected. Given the timing of the LPP waveform, this interpretation does not necessarily rule out the potential role of voluntary processes in the modulation of positive responding for later-occurring (i.e., > 500 ms) behaviors such as intentional eye movements and decision making (Issacowitz et al., 2009; Löckenhoff & Carstensen, 2007). On the other hand, it could be argued that the verbal framing task assigned to participants here was relatively easy, especially considering that the task likely became highly over-practiced during the 450 trials. This leaves open the possibility that older adults yet had sufficient cognitive resources available here to implement their presumed long-term motivation to regulate their emotions even during task performance (Isaacowitz et al., 2009; Knight et al., 2007; Mather & Carstensen, 2005). Reed and Carstensen (2012) contend that older adults who most consistently exhibit positivity effects are those with the greatest cognitive reserves available, and indeed, individuals participating in the present study were screened for any signs of cognitive impairment. Further studies with other types of task manipulation, including for example explicit instructions to modify emotion regulation goals, will likely be required to determine whether the age-related positivity effect in LPP responding is malleable or fixed.