Reduced P300 amplitude is consistently associated with trait anhedonia across repeated assessments

Abstract

Understanding how event-related potentials (ERPs) change following repeated assessments is critical to advance our understanding of neural mechanisms implicated in psychopathology. Specifically, it is unclear if associations between ERPs and individual differences can be reproduced when repeatedly measured within the same participants, or if clinical characteristics impact ERP trajectories over repeated assessments. The present study assessed P300 amplitude and latency from a flanker task at four time points over one month (M = 7.24days between assessments [SD = 1.02]) in 79 female undergraduates to examine how P300 amplitude/latency changes across repeated assessments, the presence of associations between within-and between-subjects measures of current depressive symptoms and the P300, and if between-subjects depressive symptoms moderated change in P300 over repeated assessment. Results using multilevel modeling indicated a significant reduction in P300 amplitude and latency across assessments. Individuals with increased trait anhedonia (i.e., between-subjects) exhibited reduced P300 amplitudes across assessments; there were no associations of within-subjects fluctuations in anhedonia symptoms and P300 amplitude across assessments. There was also no interaction between number of assessments and between-subjects anhedonia in relation to P300 amplitude. Unlike anhedonia, between-subjects dysphoria was unrelated to P300. These results demonstrate a relatively specific and consistent association between an attenuated P300 amplitude and trait anhedonia across repeated assessments – data that may further suggest that flanker P300 amplitude reflects hedonic and motivational processes.

1 |. INTRODUCTION

Depression is one of the most common mental health issues in the United States with a recent survey reporting that 13.1% of adults between the ages 18 and 25 experienced a current depressive disorder in 2017 (Center for Behavioral Health Statistics and Quality, 2018). The two core symptoms of major depressive disorder are depressed mood (i.e., feeling dysphoric, sad, or empty) and anhedonia (i.e., reduced pleasure or interest in almost all activities). Indeed, the presence of at least one of these two symptoms is required for a clinical diagnosis (American Psychiatric Association, 2013). The burden associated with depression is quite high: individuals with depression are more likely to be unemployed and earn a lower salary as a result of increased absenteeism and underperformance in the workplace; they are also at an increased risk to attempt suicide (Bostwick & Pankratz, 2000; Ösby et al., 2001; Whooley et al., 2002). Moreover, depression remission is difficult to achieve as past studies have reported high rates of relapse for depressive disorders (i.e., 60% after five years; Hardeveld et al., 2010). Therefore, it is critical that researchers unveil reliable measures that can increase our understanding of the etiology of the disorder, as well as improve our precision in diagnosing and treating depression. Currently, our understanding of the underlying neurophysiological mechanisms that may precede or maintain depression are not particularly well understood.

Although functional magnetic resonance imaging (fMRI) has been used extensively to study neural activity implicated in depression (Keren et al., 2018), recent empirical and meta-analytic work on psychometric properties suggests that classic task-based fMRI activations may not be suitable for individual differences (Elliott et al., 2020). On the other hand, event-related potentials (ERPs), recorded via the electroencephalogram (EEG), are direct measures of brain activity elicited by specific events or stimuli with excellent psychometric properties that can be utilized for such clinical research, including the identification of risk for psychopathology and the prediction of mental health outcomes (Hajcak et al., 2019).

The P300 is a positive-going ERP that arises approximately 300 ms after stimulus presentation and is thought to reflect processes such as attentional allocation, evaluative processing, memory, and context updating (Hajcak & Foti, 2020; Polich, 2012). P300 amplitude has been found to be robustly reduced among depressed adults (Bruder et al., 2009; Bruder et al., 2012; Gangadhar et al., 1993; Klawohn et al., 2020; Nan et al., 2018; Roschke & Wagner, 2003; Santopetro, Brush, Burani, et al., 2021; Urretavizcaya et al., 2003; Zhou et al., 2019), and has been shown to prospectively predict increases in depressive symptoms in adults with clinical depression (Santopetro et al., 2021) and female adolescents (Santopetro et al., 2020). Overall, this research indicates reductions in processing target stimuli in experimental tasks, which is consistent with broader cognitive impairments common in depressive disorders (American Psychiatric Association, 2013; Bruder et al., 2012; Clark et al., 2009; Roiser et al., 2009).

Similar to the P300 amplitude and depression literature, researchers have found that P300 latency is significantly delayed in depression - suggesting that P300 latency may also reflect an indicator of general cognitive impairments related to depression (Himani et al., 1999; Kalayam et al., 1998; Tripathi et al., 2015). For instance, Vandoolaeghe et al. (1998) reported that currently depressed adults were characterized by prolonged P300 latency compared to controls. Moreover, when examining relationships between P300 latency and specific facets of depression within currently depressed individuals, the authors found that those reporting cognitive deterioration—assessed via neuropsychological testing—exhibited longer P300 latency compared to those without cognitive deterioration. Additionally, researchers have posited that individuals characterized by increased psychomotor retardation, such as seen in melancholic depression, are more likely to exhibit prolonged P300 latency (Bruder et al., 2012). It is also important to note that some previous research has not found differences in P300 latency between currently depressed individuals and controls when elicited from auditory discrimination tasks (Blackwood et al., 1987; Roth et al., 1981).

A major limitation of most prior studies investigating the relationship between depression and P300 is that these studies consist of a single, cross-sectional assessment. To our knowledge, no previous research has examined the reproducibility of the association between P300 and depressive symptoms over time; that is, whether P300 amplitude or latency can consistently serve as a neural correlate of current depressive symptoms across multiple assessments within the same individuals. Instead, past studies have investigated repeated assessment and change in P300 amplitude in healthy samples–and findings suggest that P300 amplitude tends to decrease as a participant’s exposure to experimental stimuli becomes more frequent, which most likely reflects changes in attention allocation as participants habituate to the stimuli (Courchesne, 1978; Kinoshita et al., 1996; Polich, 1989; Wesensten et al., 1990). Thus, one possibility is that individuals experiencing depression may habituate to stimuli more rapidly, which would be reflected in a greater reduction in P300 amplitude over repeated assessments. Investigating this specific research question requires multiple testing sessions within the same individuals.

Currently, our understanding of the exact role of the P300 amplitude in depression is limited insofar as past research has specifically linked reductions in P300 amplitude to a constellation of depressive symptoms or characteristics. For example, past studies have reported that P300 amplitude is blunted in relation to increased anhedonia (Ancy et al., 1996; Gangadhar et al., 1993; Santopetro et al., 2020; Santopetro, Brush, Burani, et al., 2021; Urretavizcaya et al., 2003), increased psychotic symptoms (Karaaslan et al., 2003; Kaustio et al., 2002), heightened suicidal behavior (Hansenne et al., 1996), or greater overall depression symptom severity (Gangadhar et al., 1993; Nan et al., 2018). Thus, it remains unclear whether P300 amplitude is a general neural correlate of broad deficits related to depressive disorders, or if it relates to specific facets of depression. In the current study, we specifically examined whether P300 amplitude was associated with anhedonia and dysphoria to further examine the potential specificity of P300 reductions to either of these core symptoms of depression.

We also utilized multilevel modeling to investigate how P300 amplitude and latency changes across four EEG assessments, each separated by approximately one week. Moreover, we assessed whether there were associations between within-subject variation in core symptoms of depression and P300 amplitude/latency across assessments, and whether there were between-subject differences in core symptoms of depression in P300 across assessments. We also examined whether between-subject differences in core symptoms of depression moderated change in P300 across assessments. Lastly, we sought to examine the psychometric properties (i.e., internal consistency and test–retest reliability) of flanker P300 amplitude and latency over the month. Previous studies have demonstrated that both P300 amplitude and latency exhibit acceptable psychometrics when elicited from oddball tasks (Fabiani et al., 1987; Segalowitz & Barnes, 1993; Sklare & Lynn, 1984). Based on previous research, we hypothesized that increased depressive symptoms would be associated with reduced P300 amplitudes and prolonged P300 latencies at each assessment. Furthermore, we believed that flanker P300 amplitude would exhibit significant decreases across each assessment, as participants habituate to the stimuli and overall task. We did not have any directional hypotheses related to the flanker P300 and specific depressive symptoms, or regarding the potential interaction between depressive symptoms and changes in P300 across multiple assessments.

2 |. METHOD

The current study is part of a larger within-subject longitudinal project examining the effects of the menstrual cycle on neural activity across four repeated assessments in female undergraduates from Florida State University (FSU). The findings regarding the primary aims of this project will be reported elsewhere; the current study reports on the effects of repeated within-subject assessments on the flanker P300 and if this relationship is moderated by current depressive symptoms (i.e., anhedonia and dysphoria). All study procedures of the current study and the larger project were approved by the Institutional Review Board at FSU.

2.1 |. Participants and procedures

A total of 79 female participants were initially recruited via the university’s undergraduate psychology subject pool to participate in the project for class credit, monetary compensation, or a combination of the two. Inclusion criteria consisted of participants that were biologically female and between the ages of 18 and 35. Exclusion criteria included: (a) self-reported presence of a significant medical disability; (b) the use of hormonal contraceptives in the prior 6 months; (c) pregnancy or lactation in the prior 12 months; (d) self-reported endorsement of having an irregular menstrual cycle (i.e., participants were asked to indicate whether their menstrual cycle was nearly always regular, fairly regular, or irregular).

Participants were asked to attend a total of four assessments over the course of a month, with each visit coinciding with a specific menstrual cycle phase (mean number of days between visits was 7.24; SD = 1.02; range = 5–11). At each assessment, participants completed self-report questionnaires, biological samples (e.g., hair samples, saliva samples), and a battery of computerized, counterbalanced experimental tasks while electroencephalogram (EEG) was recorded. The order of menstrual cycle phase (i.e., which phase each participant began the study in) was counterbalanced across the four assessments. This current study utilized stimulus-locked EEG data collected during a computerized flanker task and self-report data from participants that completed any number of the assessments.

Of the initial 79 participants, 16 participants had EEG data removed from at least one assessment due to poor EEG data quality based on visual inspection, and 3 participants had EEG data removed from one assessment due to poor performance on the flanker task (i.e., error rates more than 3 SDs from the mean) at that specific visit.¹ The average exact age of our sample at baseline (N = 79) was 19.47 years old (SD = 1.84; range = 18–30), and the majority of our participants identified as White (58.4%), with the remainder identifying as Hispanic (26.0%), African-American (9.1%), Asian (3.9%), or identifying as more than one race (2.6%). Two participants did not disclose their ethnicity or race.

2.2 |. Measures

2.2.1 |. Personality inventory for DSM-5 (PID 5): Anhedonia subscale

The anhedonia subscale of the PID 5 consists of eight items (e.g., “I don’t get as much pleasure out of things as others seem to”, “Nothing seems to interest me very much”) rated on a four-point Likert scale, and was administered to participants at each study visit; there was no specific time frame in which participants were asked to rate these symptoms. Past research involving this subscale of the PID 5 suggests good internal consistency (Cronbach’s α ranging from .87–.89) in various adult samples (Fossati et al., 2013; Quilty et al., 2013). In line with these past studies, the PID 5 anhedonia subscale scores in the present sample demonstrated good internal consistency according to Cronbach’s α at each assessment (T1: .89, T2: .87, T3: .78, and T4: .76). Additionally, PID 5 anhedonia scores demonstrated good-to-excellent test–retest reliability after one week (r = .91), two weeks (r = .88) and three weeks (r = .86). An average anhedonia score was computed utilizing individual anhedonia scores from each assessment and this variable was utilized in the multilevel models.

2.2.2 |. Inventory of depression and anxiety symptoms (IDAS): Dysphoria scale

The IDAS assesses multiple, specific symptoms related to depression and anxiety disorders (Watson et al., 2007). Participants were asked to rate their mood over the past two days, rather than the past two weeks at each assessment. In the current study, we focused on the dysphoria general scale of the IDAS, which consists of 10 items that measure the core emotional and cognitive symptoms of depression such as down mood, worry, worthlessness, guilt, and psychomotor retardation or agitation. The dysphoria scale demonstrates good convergent validity with other self-reported measures of depression (i.e., Beck’s Depression Inventory-II, [BDI-II]; Beck et al., 1996), and exhibits good-to-excellent internal consistency (Cronbach’s α ranging .86–.90) in a variety of different samples (e.g., college students, clinical/psychiatric samples, high school students; Watson et al., 2007). In the present study, the IDAS dysphoria scale exhibited good-to-excellent internal consistency at each assessment (T1: .91, T2: .90, T3: .91, and T4: .86). Additionally, IDAS dysphoria scores demonstrated acceptable-to-good test–retest reliability after one week (r = .76), two weeks (r = .58) and three weeks (r = .67). An average dysphoria score was computed utilizing the individual dysphoria symptoms from each assessment and was included in the multilevel models.

2.3 |. Procedures

2.3.1 |. Flanker task

Participants completed a computerized, arrowhead version of the flanker task at each assessment. Presentation software (Neurobehavioral Systems, Albany, California) was utilized to administer the experimental task. At the start of each trial, the participant quickly viewed five horizontal arrowheads for 200 ms. The inter-trial interval (ITI) varied from 2300 ms to 2800 ms. Half of the trials would display congruent flankers (<<<<<, >>>>>), while the other half of trials displayed incongruent flankers (<<> < <, >><>>). This order was random for each participant. Participants were instructed to try their best to respond as quickly and accurately as possible to the direction of the middle arrow using their computer mouse. Participants were told to click either the left or right mouse button corresponding to the direction of the middle arrow.

Before starting the experimental task, all participants completed a 10-trial practice block. Participants had to respond correctly to at least 70% of the practice trials, or they would be asked to retake the practice block until reaching this threshold in order to assure competence. The experimental task consisted of 11 blocks of 30 trials (330 total). Participants received feedback on screen at the end of each block which was dependent on how well they were recently performing in that most recent block. If accuracy was too high (i.e., 90% or above), they would receive the message, “Please try to respond faster.” If accuracy was too low (i.e., 75% or below), they would receive the message, “Please try to be more accurate.” Otherwise, they received the message, “You’re doing a great job.” Error rates for this sample were calculated as the percentage of errors in relation to total responses for each participant and were calculated for each assessment.

2.3.2 |. EEG recording & processing

Continuous EEG was recorded during the computerized flanker task using a 32-channel active ActiCap slim electrode (Ag/AgCl) actiCHamp system (Brain Products GmbH, Gilching, Germany) at each assessment. The electrodes were situated on the scalp in accordance with the international 10/20 system electrode system, and electrode site Cz served as the online recording reference. Additional electrodes were placed behind both ears directly on the left and right mastoids, as well as electrodes placed above and below the left eye, and one to the sides of each eye (i.e., at the left and right external canthus) to collect electro-oculogram (EOG) data. Data were digitized at a sampling rate of 1000 Hz utilizing a low-pass online filter set at 100 Hz.

All EEG analyses were performed offline using Brain Vision Analyzer (version 2.2; Brain Products, Gilching, Germany). First, the continuous EEG data were re-referenced to averaged mastoids and then filtered using a 4th-order zero-phase shift Butterworth filter with cutoff frequencies of 0.01 and 30 Hz and a 24 dB/octave roll-off. Next, data were segmented in 1000 ms epochs starting 200 ms before the presentation of the flanker stimuli to 800 ms after presentation only on trials in which participants responded correctly to the direction of the middle arrow. Then, the Gratton et al. (1983) procedure was employed to correct ocular artifacts on all channels. Last, data segments containing artifacts were rejected if any data exceeded (a) a voltage step greater than 50 μV, (b) a voltage difference of 175 μV within a 400-ms interval, or (c) a maximum voltage difference of less than 0.50 μV within 100-ms intervals.

2.3.3 |. P300 quantification

Data segments were baseline corrected using the 200 ms pre-stimulus interval and averaged across all correct trials (collapsed across congruent and incongruent trials). P300 amplitude was quantified as the mean amplitude at electrode site Pz from 300 ms to 600 ms following flanker stimuli presentation, which was based on an a priori decision and is consistent with past publications utilizing this ERP component (e.g., Klawohn et al., 2020). To determine P300 peak latency, peak detection was employed to identify the most positive deflection in the 300 ms to 600 ms window post-stimulus presentation.

2.4 |. Statistical analyses

All data analyses were performed in R version 4.0.5 (R Core Team, 2021). We first conducted bivariate correlations (Pearson’s r) to examine associations between flanker P300 amplitude/latency at each assessment and self-report measures (i.e., PID 5 anhedonia and IDAS dysphoria). Next, multilevel models (MLMs) were conducted to evaluate the trajectory of change in P300 amplitude and latency across assessments, and to determine whether there were between- and within-subjects associations between anhedonia or dysphoria and P300 amplitude and latency in separate models. These MLMs also assessed whether there were between-subjects differences in change in P300 amplitude/latency across assessments as a function of anhedonia or dysphoria in separate models. In all models, within-subject P300 amplitude or latency were the dependent variables. At level 1 (repeated observations), assessment (centered at the first assessment; assessment (a) = 0, 1, 2, 3) was included in every model, and anhedonia (person-mean centered) and dysphoria (person-mean centered) were included as predictors in separate models. At level 2 (each participant), separate models included anhedonia (grand-mean centered) and dysphoria (grand-mean centered). All models allowed the intercept to vary by participant; however, to determine whether assessment should be included as a fixed or random slope, we used an empirical approach using likelihood ratio tests to determine the most appropriate and parsimonious model (Volpert-Esmond et al., 2021).² This resulted in assessment being included as a fixed slope in all models; all other variables were considered fixed effects. For all models, Satterthwaite approximations were used to estimate degrees of freedom and to obtain two-tailed p-values and maximum likelihood estimation was used to estimate model parameters. The ‘lme4’ (Bates et al., 2014) and ‘lmerTest’ (Kuznetsova et al., 2017) R packages were used to fit the MLMs. Model statistics were extracted and reported using the ‘sjPlot’ (Lüdecke, 2021) R package. The criterion of statistical significance for all analyses was p < .05.

3 |. RESULTS

3.1 |. P300 amplitude and latency correlations

Average anhedonia symptoms and average dysphoria scores were moderately associated with each other, r(74) = .41, p < .001. Participants experiencing heightened anhedonia symptoms on average exhibited significantly reduced P300 amplitudes at each assessment, r(62–73)’s ranging −.28 to −.38, ps < .016. Flanker congruency did not affect the relationship between P300 amplitude and average anhedonia symptoms: increased anhedonia symptoms at each assessment were significantly associated with reduced P300 amplitude on both congruent trials at each assessment, r(62–73)’s ranging −.30 to −.41, ps < .009, and incongruent trials at each assessment, r(62–73)’s ranging −.25 to −.32, ps <.032. However, there was no association between average dysphoria scores and P300 amplitude at any assessment. Furthermore, P300 latencies were not associated with average anhedonia or dysphoria symptoms at any assessment. More information regarding the descriptive statistics and correlations for these variables at each assessment can be found in Tables 1 and 2.

TABLE 1.

Descriptive statistics for ERP and self-report measures at each assessment

Variable	Mean	SD
T1 P300 Amplitude (μV)	9.81	4.44
T2 P300 Amplitude (μV)	8.57	4.17
T3 P300 Amplitude (μV)	7.62	4.57
T4 P300 Amplitude (μV)	6.87	4.12
T1 P300 Latency (ms)	435	61
T2 P300 Latency (ms)	407	50
T3 P300 Latency (ms)	409	57
T4 P300 Latency (ms)	392	56
T1 Anhedonia	7.28	4.79
T2 Anhedonia	7.06	4.98
T3 Anhedonia	7.35	4.89
T4 Anhedonia	6.44	4.67
T1 Dysphoria	20.33	8.25
T2 Dysphoria	20.66	8.42
T3 Dysphoria	20.67	8.60
T4 Dysphoria	19.69	6.73

Notes: T1 represents baseline assessment (n = 75), T2 represents second assessment (n = 74), T3 represents third assessment (n = 75), and T4 represents fourth assessment (n = 64). Anhedonia was measured utilizing PID 5 and dysphoria symptoms were measured from the IDAS.

Table 2.

Correlations between anhedonia and dysphoria scores with flanker P300 amplitude/latency at each assessment

Variable	T1 Anh.	T2 Anh.	T3 Anh.	T4 Anh.	T1 Dys.	T2 Dys.	T3 Dys.	T4 Dys.
T1 P300 Amplitude (μV)	−.30**	−.26*	−.29*	−.27*	−.19	−.16	−.18	−.05
T2 P300 Amplitude (μV)	−.25*	−.27*	−.26*	−.24*	−.07	−.05	−.09	.04
T3 P300 Amplitude (μV)	−.32**	−.35**	−.33**	−.25*	−.17	−.09	−.21	.03
T4 P300 Amplitude (μV)	−.36**	−.38**	−.43***	−.32*	−.20	−.17	−.33**	−.03
T1 P300 Latency (ms)	−.08	−.01	−.03	−.04	−.15	−.05	−.09	−.03
T2 P300 Latency (ms)	.06	.05	.07	−.02	−.14	−.19	−.07	−.12
T3 P300 Latency (ms)	−.02	.05	.00	−.08	−.12	−.15	−.17	−.11
T4 P300 Latency (ms)	−.01	−.02	−.09	−.09	.05	.08	.10	.04

Note: T1 represents baseline assessment (n = 75), T2 represents second assessment (n = 74), T3 represents third assessment (n = 75), and T4 represents fourth assessment (n = 64). Anhedonia (Anh.) was measured utilizing PID 5 and dysphoria (Dys.) symptoms were measured from the IDAS.

^*Indicates p <.05;;

^**Indicates p <.01;;

^***Indicates p <.001.

3.2 |. P300 MLMs

3.2.1 |. P300 amplitude

For the unconditional model, the mean intercept (b = 8.18, SE = 0.45, t[79.03] = 18.3, p <.001) and the variance of the intercept across individuals (Likelihood ratio test:

χ² [1] = 190.39, p <.001) was significant. The ICC was 0.726, suggesting approximately 72.6% of the variance in P300 amplitude was accounted for between individuals, while 27.4% was accounted for within individuals. The unconditional model indicated a linear pattern of decline in P300 amplitude across assessments. Therefore, the models predicting P300 amplitude were specified as:

• $\mathit{\text{Model}}3:P300\mathit{\text{Latency}}=\mathit{\text{Assessment}}*\mathit{\text{Anhedonia}}\left(\mathit{\text{BS}}\right)+\mathit{\text{Anhedonia}}\left(\mathit{\text{WS}}\right)+\left(1\right|\mathit{\text{ID}})$

• $\mathit{\text{Model}}2:P300\mathit{\text{Latency}}=\mathit{\text{Assessment}}*\mathit{\text{Dysphoria}}\left(\mathit{\text{BS}}\right)+\mathit{\text{Dysphoria}}\left(\mathit{\text{WS}}\right)+\left(1\right|\mathit{\text{ID}})$

Note. “BS” represents between-subjects and “WS” represents within-subjects.

In the first conditional model (Model 1), we examined associations between within-subject and between-subject anhedonia with P300 amplitude and whether there were between-subjects differences in change in P300 amplitude across assessments. The full model is reported in Table 3. The main effect of assessment was significant consistent with a linear decrease in P300 amplitude across assessments. There was also a significant main effect of between-subjects anhedonia, confirming that individuals with heightened anhedonia scores on average exhibited reduced P300 amplitudes. However, there was no interaction between number of assessments and between-subjects anhedonia.³ Lastly, there was no association between within-subjects anhedonia symptoms and P300 amplitude across assessments. The two main effects of assessment and between-subjects anhedonia are depicted in Figures 1 and 2, respectively. Figure 1 presents change in flanker P300 amplitude across assessments at varying levels of between-subjects anhedonia. Figure 2 illustrates the stimulus-locked grand average waveforms comparing high versus low anhedonic participants at each assessment according to a median-split of PID 5 anhedonia scores.

TABLE 3.

Multilevel models 1 (left) and 3 (right) involving anhedonia scores and assessment to predict flanker P300 amplitude and latency

	Amplitude			Latency
Predictors	Estimates	CI	P	Estimates	CI	P
(Intercept)	9.50	8.62–10.39	<.001	430.33	417.86–442.79	<.001
Anhedonia between	−0.27	−0.46 - −0.07	.007	−0.23	−2.93 – 2.48	.868
Assessment	−0.92	−1.13 - −0.71	<.001	−10.98	−14.58 - −7.38	<.001
Anhedonia within	0.05	−0.11 – 0.21	.522	−1.35	−4.13 – 1.43	.339
Anhedonia between x Assessment	−0.03	−0.07 – 0.02	.249	−0.10	−0.86 – 0.66	.802
Random Effects
σ2	3.60			1070.61
τ00	12.80ID			2271.05ID
ICC	0.78			0.68
N	77ID			77ID
Observations	279			279
Marginal R2/Conditional R2	0.156/0.815			0.042/0.693

FIGURE 1.

Slopes of flanker P300 amplitude across assessments at varying levels of between-subjects anhedonia (model 1). Assessment number is scaled 1 through 4 in the above plot which differs compared to how assessments were scaled in the MLMs (i.e., 0 through 3).

FIGURE 2.

Flanker stimulus-locked waveforms of high anhedonic participants (n = 32) compared to low anhedonic participants (n = 32) at each assessment (a = first, b = second, c = third, d = fourth).

We re-conducted these analyses and included age, menstrual cycle phase order, and days between assessments to the model. Overall, the results were consistent with Model 1. The main effect of assessment, b = −0.92, t(204.69) = −8.65, p <.001 and between-subjects anhedonia, b = −0.26, t(95.28) = −2.65, p = .010, remained. There were no additional main effects of age, phase order, or number of days between assessments (all ps>.176). There was also no association between within-subject anhedonia and P300 across assessments, and there was no interaction between assessment and between-subjects anhedonia (both ps>.248).

Additionally, the current MLM results are consistent when predicting flanker P300 amplitude elicited from congruent or incongruent trials. For congruent trials only, number of assessments, b = −1.00, t(205.04) = −8.63, p < .001, and between-subjects anhedonia, b = −0.26, t(101.10) = −2.83, p = .006, were significant predictors of P300 amplitude, whereas there was no significant interaction between assessment and between-subjects anhedonia, nor was within-subjects anhedonia a significant predictor (all ps>.240). For incongruent trials only, assessment number, b = −0.85, t(204.10) = −7.31, p < .001, and between-subjects anhedonia, b = −0.29, t(92.78) = −2.56, p = .012, were significant predictors of P300 amplitude, and there was again no significant interaction or significant main effect of within-subjects anhedonia (all ps>.480).

In the second conditional model (Model 2), we examined associations between within-subject and between-subject dysphoria with P300 amplitude and whether there were between-subjects differences in change in P300 amplitude across assessments. There was a significant main effect of assessment on P300 amplitude consistent with a linear decrease in P300 amplitude across assessments. However, no other main effects or interactions involving between-subjects or within-subjects dysphoria scores were significant. More information regarding this model is presented in Table 4.

TABLE 4.

Multilevel models 2 (left) and 4 (right) involving dysphoria scores and assessment to predict flanker P300 amplitude and latency

	Amplitude			Latency
Predictors	Estimates	CI	P	Estimates	CI	P
(Intercept)	9.47	8.53–10.40	<.001	431.43	419.04–443.82	<.001
Dysphoria between	−0.06	−0.20 – 0.08	.419	−1.44	−3.27 – 0.39	.123
Assessment	−0.92	−1.12 – −0.71	<.001	−11.36	−14.89 – −7.82	<.001
Dysphoria within	−0.00	−0.06 – 0.05	.881	−1.20	−2.15 – −0.26	.013
Dysphoria between x Assessment	−0.02	−0.05 – 0.01	.282	0.10	−0.41 – 0.61	.703
Random Effects
σ2	3.54			1045.12
τ00	14.54id			2222.60ID
ICC	0.80			0.68
N	76id			76id
Observations	275			275
Marginal R2/Conditional R2	0.071/0.818			0.073/0.703

3.2.2 |. P300 latency

The effects of between- and within-subjects anhedonia/dysphoria scores and multiple assessments on flanker P300 latency were also investigated. For the unconditional model, the mean intercept (b = 413.49, SE = 5.43, t[78.37] = 76.16, p<.001) and the variance of the intercept across individuals (Likelihood ratio test: χ² [1] = 105.98, p <.001) was significant. The ICC was 0.572, suggesting approximately 57.2% of the variance in P300 latency was accounted for between individuals, while 42.8% was accounted for within individuals. The unconditional model indicated a linear pattern of decline in P300 latency across assessments. Therefore, the models predicting P300 latency were specified as:

• $\mathit{\text{Model}}3:P300\mathit{\text{Latency}}=\mathit{\text{Assessment}}*\mathit{\text{Anhedonia}}\left(\mathit{\text{BS}}\right)+\mathit{\text{Anhedonia}}\left(\mathit{\text{WS}}\right)+\left(1\right|\mathit{\text{ID}})$

• $\mathit{\text{Model}}4:P300\mathit{\text{Latency}}=\mathit{\text{Assessment}}*\mathit{\text{Dysphoria}}\left(\mathit{\text{BS}}\right)+\mathit{\text{Dysphoria}}\left(\mathit{\text{WS}}\right)+\left(1\right|\mathit{\text{ID}})$

Note. “BS” represents between-subjects and “WS” represents within-subjects.

In the conditional model including both between- and within-subjects anhedonia scores (Model 3) results revealed a significantly reduced (i.e., shorter) P300 latency across assessments. No other main effects or interactions including within-subjects or between-subjects anhedonia symptoms were significant. More information regarding this model is presented in Table 3.

In the conditional model including both between-and within-subjects dysphoria (Model 4) results again revealed a significant linear decline in P300 latency across assessments. There was no main effect of between-subjects dysphoria symptoms or interaction between assessment and between-subjects dysphoria. However, there was a significant main effect of within-subjects dysphoria scores, suggesting that during assessments where participants experienced increased dysphoria compared to their average, their P300 latency tended to be shorter. More information regarding this model is presented in Table 4.

3.3 |. Psychometric properties of the P300

3.3.1 |. Internal consistency

To measure internal consistency of P300 amplitude and latency, we first correlated P300 amplitude/latency on odd- and even-numbered trials (e.g., Levinson et al., 2017), and then corrected the correlation coefficients using the Spearman-Brown prophecy formula (Nunnally et al., 1967). P300 amplitude exhibited excellent internal consistency at each assessment (T1: .98, T2: .98, T3: .96, and T4: .99), while P300 latency exhibited good-to-excellent internal consistency at each assessment (T1: .84, T2: .88, T3: .82, and T4: .92). The baseline assessment (T1) consists of 77 participants, T2 consists of 75 participants, T3 consists of 76 participants, and T4 consists of 64 participants.

3.3.2 |. Test–retest reliability

Test–retest reliability was computed for flanker P300 amplitude and latency over the course of one week (i.e., baseline to second assessment), two weeks (i.e., baseline to third assessment), and three weeks (i.e., baseline to fourth assessment). Flanker P300 amplitude demonstrated acceptable-to-good test–retest reliability after one week (r = .82), two weeks (r = .78) and three weeks (r = .79). Flanker P300 latency exhibited adequate test–retest reliability over one week (r = .62), two weeks (r = .63), and three weeks (r = .53). One week reliability consists of 74 participants, two-week reliability consists of 74 participants, and three-week reliability consists of 63 participants.

4 |. DISCUSSION

A major aim of the present study was to investigate naturalistic changes in P300 amplitude and latency over multiple assessments. In line with past studies that assessed P300 amplitude across repeated assessments in healthy samples (Courchesne, 1978; Kinoshita et al., 1996; Polich, 1989; Wesensten et al., 1990), we found that P300 amplitude significantly decreased with each additional EEG assessment. P300 latency also decreased across repeated assessments. Together, these results suggest that habituation to experimental stimuli (i.e., both congruent and incongruent flanker stimuli) via repeated assessment may ease processing of the target stimuli and therefore steadily reduce the individual’s deployment of cognitive resources in response to more familiar stimuli, resulting in decreased P300 amplitude and shorter P300 latency.

Next, we sought to examine the nature of the relationships between current depressive symptoms, both between- and within-subjects effects, and P300 amplitude and latency across repeated assessments. Although average anhedonia and dysphoria symptoms were moderately interrelated (i.e., r = .41), reduced flanker P300 amplitude was only consistently associated with heightened between-subjects anhedonia symptoms; P300 amplitude was unrelated to elevated between- or within-subjects dysphoric mood or fluctuations in within-subject anhedonia symptoms. This finding was also independent of factors such as menstrual cycle phase which has been theorized to significantly affect mood (Eriksson et al., 2002).

A handful of studies have previously observed that reduced P300 amplitude is specifically associated with increased anhedonia in individuals with heightened depression (Ancy et al., 1996; Gangadhar et al., 1993; Santopetro et al., 2020; Santopetro, Brush, Burani, et al., 2021; Urretavizcaya et al., 2003) suggesting that deficits in motivational processing might be especially relevant to the assoication between P300 and depression. The current findings extend this work to further demonstrate that deficits in P300 amplitude may reflect more trait-like features of depression, specifically anhedonia, as P300 amplitude did not appear to relate to within-subject fluctuations in concurrent depressive symptoms in the present study. Lastly, these novel findings demonstrate that the relationship between P300 amplitude and anhedonia is quite robust and reproducible within the same sample over multiple (i.e., at least four) assessments.

Conversely, there were no associations involving between-subjects depressive symptoms (i.e., anhedonia or dysphoria) and P300 latency at any assessment, which contrasts with literature suggesting that longer P300 latency is characteristic of depressive disorders (Himani et al., 1999; Kalayam et al., 1998; Tripathi et al., 2015; Vandoolaeghe et al., 1998). Indeed, the current study found some evidence that higher within-subjects dysphoria, not anhedonia, was related to shorter P300 latency—a finding that contrasts with previous literature. It is possible that P300 latencies measured in different experimental tasks across studies are sensitive to different processes, as the current study used the flanker task, while most previous researchers have used auditory oddball tasks. Moreover, it is possible that prolonged P300 latencies may only characterize individuals experiencing more severe (i.e., clinical) levels of depression or those individuals experiencing increases in psychomotor retardation or melancholia (Bruder et al., 2012).

We tested whether current depressive symptoms, on average (i.e., between-subjects effect), moderated change in both P300 amplitude and latency across four assessments. Results indicated a lack of moderation, suggesting that changes in P300 amplitude/latency across assessments were relatively stable regardless of between-subjects anhedonic or dysphoric mood symptoms. This was further evidenced by the consistent correlations between anhedonia scores and reduced P300 amplitude at each of the four assessments (see Table 2). These findings suggest that the underlying neurophysiological mechanisms associated with reductions in P300 amplitude across repeated assessments may be largely separate from pathways that impact P300 amplitude in relation to depression.

A final aim of the present study was to examine the psychometric properties of the flanker P300 across multiple assessments; more specifically, we examined the internal consistency and test–retest reliability of P300 amplitude and latency. Past studies have investigated the psychometric properties of the P300 (amplitude and latency) elicited from oddball tasks and reported that the P300 exhibits acceptable within- and between-subjects reliability (Fabiani et al., 1987; Segalowitz & Barnes, 1993; Sklare & Lynn, 1984). The present study extends this critical psychometric work to the P300 component measured via the flanker task to address recent calls for establishing whether the neurophysiological measures employed are reliable enough to understand mental health disorders (Clayson et al., 2021; Clayson & Miller, 2017; Elliott et al., 2020; Hajcak et al., 2017). One major concern is that neurophysiological measures with poor psychometric properties may be too unreliable to make valid inferences regarding individual differences. In the current study, flanker P300 amplitude consistently demonstrated excellent internal reliability at each time point (.96 to .99) which is directly in line with other recent research from our group (see Klawohn et al., 2020; Santopetro et al., 2020; Santopetro, Brush, Bruchnak, et al., 2021; Santopetro, Kallen, Threadgill, et al., 2021). Flanker P300 amplitude also exhibited good test–retest reliability after one week (r = .82), two weeks (r = .78) and three weeks (r = .79). Past research has demonstrated that peak measures of ERPs are typically less reliable compared to mean measurements (e.g., Clayson et al., 2013); however, flanker P300 latency quantified as a peak measure exhibited good internal consistency (.82 to .92) and adequate test–retest reliability over the course of the study. Despite these findings, it is important to note that peak measures of ERPs are highly susceptible to noise levels in the EEG data which may limit interpretability. In sum, both flanker P300 amplitude and latency exhibit strong psychometric properties when assessed repeatedly within the same sample, which provides support for their utility as individual difference measures in future investigations.

There were some limitations to the current study. First, the sample was comprised of only relatively young female participants, which reduces the generalizability of our findings. Further research is needed to examine if males exhibit the same changes in P300 amplitude and latency across repeated assessments, and to assess whether the P300 amplitude and anhedonia relationship is reproducible over multiple assessments in males. Second, the current study consisted of an unselected and relatively “healthy” sample of college-educated participants that exhibited low depressive symptoms, on average. Future projects should attempt to replicate these findings in large, diverse samples of individuals with clinical levels of depression.

In conclusion, the present study found reductions in P300 amplitude and faster P300 latency with repeated assessments within the same sample. Additionally, both P300 amplitude and latency exhibited good psychometric properties in terms of internal consistency and test–retest reliability. Reductions in P300 amplitude were related to increased symptoms of trait anhedonia, but not dysphoria. In this way, these data implicate motivational processes in the functionality of the P300. Indeed, the P300 amplitude and trait anhedonia relationship was consistently reproducible at each assessment within the same sample over the course of a month.

DATA AVAILABILITY STATEMENT

Data or other materials are available through correspondence with the author.