Economic Choice and Heart Rate Fractal Scaling Indicate That Cognitive Effort Is Reduced by Depression and Boosted by Sad Mood

Abstract

BACKGROUND:

People with depression typically exhibit diminished cognitive control. Control is subjectively costly, prompting speculation that control deficits reflect reduced cognitive effort. Evidence that people with depression exert less cognitive effort is mixed, however, and motivation may depend on state affect.

METHODS:

We used a cognitive effort discounting task to measure propensity to expend cognitive effort and fractal structure in the temporal dynamics of interbeat intervals to assess on-task effort exertion for 49 healthy control subjects, 36 people with current depression, and 67 people with remitted depression.

RESULTS:

People with depression discounted more steeply, indicating that they were less willing to exert cognitive effort than people with remitted depression and never-depressed control subjects. Also, steeper discounting predicted worse functioning in daily life. Surprisingly, a sad mood induction selectively boosted motivation among participants with depression, erasing differences between them and control subjects. During task performance, depressed participants with the lowest cognitive motivation showed blunted autonomic reactivity as a function of load.

CONCLUSIONS:

Discounting patterns supported the hypothesis that people with current depression would be less willing to exert cognitive effort, and steeper discounting predicted lower global functioning in daily life. Heart rate fractal scaling proved to be a highly sensitive index of cognitive load, and data implied that people with lower motivation for cognitive effort had a diminished physiological capacity to respond to rising cognitive demands. State affect appeared to influence motivation among people with current depression given that they were more willing to exert cognitive effort following a sad mood induction.

Depression is associated with cognitive control deficits, with adverse implications for regulating moods or pursuing goals in daily life (1–3). Cognitive control is effortful (4–8), prompting speculation that deficits reflect a reduced propensity to exert cognitive effort rather than reduced capacity for control (1,9–12). By one hypothesis, people with depression exert effort less because they are more sensitive to effort costs, are less sensitive to potential benefits, and expect that their attempts at control will be less efficacious (10).

However, evidence that depressed individuals are less willing to exert cognitive effort is mixed. Evidence in favor includes that more severe depression correlates with less self-reported effort (13) and that people with depression stop expending cognitive effort sooner than control subjects do in a progressive ratio task (14). People with current or remitted depression and those with higher anhedonia are also less likely to choose physical effort for rewards (15–17) and report greater effort than indicated by their actual handgrip squeeze force (18).

On the other hand, adults diagnosed with major depression choose high cognitive effort to earn money as much as control subjects, despite being less likely to choose high physical effort (19). Psychophysiological data also argue against the hypothesis that people with depression exert less cognitive effort. For example, cardiovascular indices (e.g., cardiac pre-ejection period [PEP] and systolic blood pressure) are elevated among people with depression when cognitive demands are low and are attenuated only when demands are high (20–22). Thus, people with depression may not be less willing to exert effort, but simply less able.

If people with depression are less willing to exert effort, an important question is whether lower motivation is traitlike or whether it varies by depression status. Specifically, we asked whether people with remitted depression also differed from control subjects in terms of willingness to expend cognitive effort. This question has clinical implications. In one study, decision making about physical effort predicted future relapse among individuals with remitted depression (15).

Willingness to exert cognitive effort may also depend on affect. Both sadness and amotivation are cardinal features of depression, suggesting intuitively that sadness saps motivation. On the other hand, sad affect can enhance memory and decision making among healthy adults (23,24) and those with depression (25). Also, sad primes can increase cardiac indices of engagement, suggesting that sad stimuli may promote motivation (26). Thus, while we hypothesized a priori that sad affect reduces motivation, it is conceivable that a sad mood induction could increase motivation.

Our study tested the hypotheses that motivation for cognitive effort would be diminished in depression and modulated by sad affect. We measured willingness to exert effort in terms of cognitive effort discounting (COGED) (4): participants’ propensity to choose less money for a low-load working memory task versus more money for a high-load working memory task. COGED choices vary with reward sensitivity (27–29), striatal dopamine signaling (28), and clinically relevant deficits in schizophrenia (27,30), yet they have never been used to quantify motivation in depression. Given that COGED choices correlate with effort in daily life among healthy adults (4), we also tested whether choices predicted clinician-rated global functioning across healthy and depressed individuals.

We measured on-task effort exertion by cardiovascular reactivity (31–34). Specifically, we used detrended fluctuation analysis (DFA) to test for the (loss of) heart rate fractal scaling with cognitive load. Fractal refers to similarity in complexity across large scales (e.g., at low heart rate frequencies) and small scales (high frequencies). Fractal dynamics arise from sympathetic and parasympathetic system interactions at rest, organizing similar variability across timescales (35–37). Fractal structure breaks down when vagal control is withdrawn to reallocate resources in response to external demands (32), producing more correlated or periodic behavior (37,38). Importantly, people with depression show blunted or less context-appropriate vagal withdrawal to diverse (e.g., emotional or stressful) stimuli, perhaps reflecting impaired capacity to respond to external demands (39–41) [though see (42)].

We chose to examine fractal scaling because of limitations in more widely used cardiovascular measures. Since the seminal work of Obrist (43) and Wright and Kirby (44), multiple cardiovascular dynamics have been mapped to cognitive effort. For example, shorter PEP, higher systolic blood pressure, and lower respiratory sinus arrhythmia (RSA) all are associated with cognitive load (34,45–47). Yet, whether such measures scale monotonically with subjective effort remains unclear. For example, one study showed that PEP and systolic blood pressure increase with demands up to a point, but then revert at higher demands (48), suggesting that they index task engagement as a function of ability, rather than pure subjective effort. Furthermore, while these more common cardiovascular measures can distinguish people with depression from control subjects in terms of cognitive demand (20,21,34,49), they have failed to do so in other studies (33,50).

Fractal dynamics are a promising alternative given that they vary monotonically with both physical and cognitive demand (38,51). Moreover, fractal scaling is relatively more stable within individuals (52) and is more normally distributed (53) than many cardiovascular measures and thus may be a more sensitive index of subjective changes in state. Also, in contrast to other methods, DFA is robust to nonstationarity (54).

METHODS AND MATERIALS

Adults (age range, 30.0–58.5 years) with current (n = 36) or remitted (n = 67) depression and healthy control subjects (n = 49) were recruited (demographics in Table S1) from a larger project on risk factors for juvenile-onset mood disorders [for original recruitment details see (55,56)]. Participants were chosen according to DSM-IV criteria, using Structured Clinical Interview for DSM-IV2Patient Edition (57), by trained clinical interviewers and confirmed by best estimate clinical consensus (58). We assessed participants’ functioning via the Global Assessment of Functioning scale by trained and experienced professional clinical evaluators. All procedures were approved by the Institutional Review Board at the University of Pittsburgh. Written informed consent was obtained from all participants.

Willingness to exert cognitive effort was measured by the COGED paradigm (4) (Figure 1A, B). In COGED, participants choose between offers of a hard n-back task (n = 2–4) for more money ($2.00 or $5.00) or an easy task (n = 1) for less money. The offer for 1-back for each {task, money} pair starts at half the offer for the hard task (e.g., $2.50 for the 1-back vs. $5.00 for the 3-back). After each choice, the easy task offer amount is titrated up or down, with the adjustment being half as much on each subsequent choice, until participants are indifferent between the offers for each hard task–easy task pair (after 5 choices).

Figure 1.

Paradigm overview. Cognitive effort discounting (COGED) paradigm and experimental session overview. (A) Before making COGED decisions, participants experienced 2 rounds of each level of the n-back (n = 1–4). Stimuli consisted of upper case consonants presented one at a time. Participants indicated by button press whether each stimulus was a target or nontarget. Load levels are referred to by lower-case vowels: in the “a” task (the 1-back) a stimulus was a target if repeated after 1 item in sequence, in the “i” task (the 3-back) a stimulus was a target if repeated after 3 items in sequence, etc. (B) After practicing all levels of the nback, participants made a series of choices between paired offers to repeat one of the harder levels of the n-back (n = 2–4) for more money or the easy level (n = 1) for less money. Offers for the easy level were titrated after each choice, asymptotically (5 choices for each load pairing and hard level offer amount) approaching an indifference point, which reflected the degree to which participants subjectively discounted offers based on the perceived difference in effort costs. (C) The session paradigm featured 2 unpaced breathing periods for resting heart rate dynamics; n-back practice; and, immediately following a brief sad or neutral film, COGED choices.

Indifference points define the subjective cost of cognitive effort. For example, if a participant is indifferent between $5.00 for the 3-back and $1.50 for the 1-back, the additional effort of the 3-back subjectively costs $3.50. We defined subjective value (SV) as the indifference point divided by the hard task offer (in this example $1.50/$5.00 = 0.30). Conversely, indifference points quantify willingness to exert effort: Higher SVs indicate greater willingness to choose higher effort for reward. We summarized participants’ motivation by calculating the area under the curve (AUC) defined by mean SV over all hard task load levels and offer amounts (n = 2–4 for $2 and $5).

Participants experienced all levels of the n-back task first (n = 1–4) (Figure 1A). Thus, all participants experienced the n-back under the same conditions. Next, they watched a brief either sad or neutral film, followed immediately by COGED (Figure 1B, C). Participants were asked to execute one of their choices, selected at random among all their choices, by repeating the corresponding n-back task in exchange for the offer they had chosen.

The n-back stimuli (upper-case consonants) were presented for up to 2.5 seconds, during which participants indicated by button press whether stimuli were targets or nontargets. Stimuli were followed by a fixation cross to maintain a fixed trial duration of 3.0 seconds. Participants completed 3 rounds of 40 trials of each n-back level. Feedback about accuracy was given at the end of each round.

Participants were pseudorandomly assigned to watch a 3-minute, either sad or neutral mood induction film clip. The sad film clip was from the movie The Champ, depicting a boy’s anguish over his father’s death, and has been widely used to induce sadness (59). The neutral film depicted aquatic scenes, inducing minimal physiological arousal (60).

Electrocardiography was continuously recorded using disposable, pre-gelled Ag/AgCl spot electrodes (ConMed Andover Medical) with a modified lead II configuration. Cardiac signals were sampled online at 1000 Hz using a MindWare BioNex system and BioLab software (MindWare Technologies Ltd.). Electrocardiography data were inspected, and artifacts were corrected manually by trained raters. Interbeat (R-R) intervals were then extracted by using MindWare heart rate variability analysis software version 3.1.4, which we further used to compute PEP (a measure of sympathetic activity) and high-frequency heart rate variability (an index of RSA, which reflects parasympathetic activity) during n-back task engagement; see Supplement for results.

To quantify fractal scaling, we computed the short-term scaling exponent, ${\alpha }_1$, derived via DFA (54), from interbeat (R-R) intervals (35,38,61). First, we took the cumulative sum of the mean-centered interval series, $b$:

$$y(t)=\sum _{t=1}^T{b}_t-\overline{b}$$ (1)

Next, we segmented the series into $k$ windows, of sizes ranging from 4 to 16 beats [logarithmically spaced, by convention (62,63)]. Then, the series $y(t)$ was detrended in each segment $k$ by removing the linear least-squares fit, producing $y_k(t)$. Next, the fluctuation function $F_N$ was computed for each segment of size $N$.

$$F_N=\sqrt{\frac{1}{N}\sum _{j=1}^N{(y(j)-y_k(j))}^2}$$ (2)

Finally, we estimated the exponent a1 as the slope of a log-log plot of the arithmetic mean of the fluctuation function against its corresponding segment size. DFA values were computed using the algorithm implemented in the PhysioNet Cardiovascular Signal Toolbox version 1.1.0 (62,63).

RESULTS

Participants experienced higher levels of the n-back task as more costly, progressively discounting offers to repeat harder n-back tasks (Figure 2A, B). A multilevel regression SVs on load, controlling for offer amount, n-back performance, group, condition, age, and gender, revealed a reliable effect of n-back load (b = −0.61, SE = 0.029, p < .001). Better n-back performance (sensitivity index d′) also correlated with shallower discounting (b = 0.081, SE = 0.029, p = .005), and there was a trending linear effect of increasing age on steeper discounting (b = −0.092, SE = 0.047, p = .050). While these results replicate prior findings (4), including gender as an exploratory covariate further revealed that men discounted more shallowly than women (b = 0.23, SE = 0.11, p = .030) (Table S2).

Figure 2.

Subjective values of cognitive effort–discounted offers by group and mood induction. Subjective values were computed separately for each level and hard task offer amount ($2 or $5). Subjective values are shown here, averaged across offer amounts following (A) a neutral film and (B) a sad film. (C) Individual differences in discounting quantified by area under the curve. Points show mean area under the curve ± SEM; p values reflect within-group, or within-condition, effects, controlling for age, gender, and mean n-back performance (d′).

To summarize individual differences, we computed the AUC (Figure 2C inset) capturing participants’ discounting without assuming the shape of their discounting function. A regression of AUC values controlling for age, gender, and mean n-back performance (d′) revealed that control subjects discounted less than participants with depression (b = 0.69, SE = 0.32, p = .036), while participants with remitted depression did not differ from control subjects (b = 0.22, SE = 0.29, p = .45). Steeper discounting supports the hypothesis that people with depression are less willing to exert cognitive effort.

Interestingly, discounting was shallower after a sad versus a neutral priming film (b = 1.0, SE = 0.33, p = .0018) (Figure 2C). This motivation-enhancing effect of a sad film was strongest for participants with current depression. In the regression of AUC values, group differences were captured by a negative group by mood interaction for both remitted (b = 21.0, SE = 0.43, p = .013) and control (b = 21.0, SE = 0.40, p = .021) groups versus participants with depression. Indeed, a simpler AUC regression involving only data for participants with depression revealed a clear effect of a sad versus neutral film increasing the SV of offers (b = 0.96, SE = 0.32, p = .0047). There was no corresponding mood induction effect on the SV of offers among control subjects (b = −0.0017, SE = 0.28, p = .995) or participants with remitted depression (b = −0.020, SE = 0.26, p = .94). Note that self-rated sad mood increased reliably for each group relative to preinduction ratings, and groups were indistinguishably sad following the sad film (Figure S1).

If a sad mood induction boosts motivation in depression, this should counteract group effects on effort discounting. To test this prediction, we regressed AUC on group and film type, controlling for age, gender, and n-back d′. Consistent with the prediction, control subjects discounted less than participants with depression following the neutral film (b = 0.75, SE = 0.33, p = .027), but not following the sad film (b = −0.30, SE = 0.30, p = .32). Moreover, following the sad film, participants with current depression discounted less than those with remitted depression (b = −0.77, SE = 0.30, p = .012). Hence, a brief sad film boosted motivation for people with current depression to levels that exceed motivation among those with remitted depression (Figure 2C).

Relationship Between Effort Discounting and Global Functioning

We tested the prediction that shallower discounting also implies greater propensity to exert effort to pursue goals in daily life by regressing on AUC two correlated (Pearson’s r = 0.96, 95% CI [0.95, 0.97], p < .001) but distinct, measures of social, psychological, and occupational functioning derived from the Global Assessment of Functioning scale: 1) the highest level of functioning from the past year and 2) the lowest level of functioning during the week of poorest functioning in the past month.

We separately regressed Global Assessment of Functioning scores on AUC, controlling for age, gender, mood induction, and n-back performance as well as years of education, current depression symptoms (Beck Depression Inventory-II) (64), anxiety symptoms (Beck Anxiety Inventory) (65), and medication status (whether taking any psychoactive medication or not). Regressions revealed relationships between AUC and the highest level of global functioning over the past year (b = 0.13, SE = 0.055, p = .020) (Figure 3A) and during the week of poorest functioning in the last month (b = 0.14, SE = 0.055, p = .010) (Figure 3B). In both cases, willingness to exert cognitive effort correlated with better global functioning (Tables S4, S5).

Figure 3.

Global Assessment of Functioning (GAF) correlated with effort discounting. Higher GAF correlated with shallower discounting (higher area under the curve [AUC]) for both (A) the highest functioning over the past year and (B) the lowest level functioning during the week of poorest functioning over the past month. Figures show the relationship between AUC and GAF scores after regressing out age, gender, Beck Depression Inventory-II scores, medication status, average n-back performance, and mood induction. A regression line is plotted with standard errors.

Autonomic Constraints

Heart rate fractal scaling was quantified to index on-task effort exertion via DFA (54). DFA, in brief, quantifies the increase in heart rate variability from small to large scales (${\alpha }_1$) (Figure 4A; see Methods and Materials). Fractal scaling was maximized when variability was matched across scales (${\alpha }_1\approx 1.0$) and was diminished when dynamics were either more random (${\alpha }_1<1.0$) or more correlated (${\alpha }_1>1.0$).

Figure 4.

Heart rate fractal scaling across the paradigm, quantified via detrended fluctuation analyses (DFAs). (A) Two examples of interbeat interval (IBI) series along with the corresponding log-log plot of the fluctuation function ($<F_N>$; see Methods and Materials) for each series versus the window size used to quantify the fluctuation function. The slope of the log-log plot quantified the DFA slope (${\alpha }_1$) based on window sizes ranging from 4 to 16 beats. (B) Within-participant ${\alpha }_1$ values were stable across repeated measurements (unpaced breathing during rest time 1 and 2). (C) When viewing either film across participants, ${\alpha }_1$ increased reliably. (D) Mean ± SEM of ${\alpha }_1$ increased parametrically with n-back load. In the plots, ${\alpha }_1$ values are separated by group and by median split on area under the curve across both sad and neutral film conditions.

Fractal scaling was stable within individuals across time points (across rest periods; Pearson’s r = 0.56, 95% CI [.48, .63]; p < .001) (Figure 4B). By contrast, increases in ${\alpha }_1$ indicated a similar (p = .84) loss of fractal scaling when participants viewed either sad (mean = 1.2) or neutral (mean = 1.2) films versus rest (mean = 1.0; both p < .001) (green in Figure 4C). This pattern was consistent with vagal withdrawal during film viewing (40). Importantly, there was also a clear parametric effect of n-back load on ${\alpha }_1$, which rose progressively from the 1-back through the 4-back (Figure 4D). Also, steep discounting predicted a blunting of this pattern among participants with current depression (Figure 4D). In contrast, participants with current depression who discounted shallowly showed the same fractal scaling with load as control subjects and participants with remitted depression. Neither parametric load effects nor discounting effects were apparent with either PEP or RSA (Figures S2, S3).

A blunted physiological response among steep discounters with depression suggests a relationship between autonomic reactivity and cognitive effort. To estimate the relationship between reactivity and discounting, we regressed SV on load-specific ${\alpha }_1$, load, and group, controlling for offer amount and performance (d′). The multilevel model also controlled for variables that we found to affect discounting: mood induction, age, and gender. Along with reliable effects of load (b = −0.61, SE = 0.068, p < .001), sad mood (b = 0.40, SE = 0.14, p = .0051), gender (men vs. women: b = 0.19, SE = 0.084, p = .025), and n-back performance (b = 0.062, SE = 0.024, p = .0087), there was also a positive effect of DFA ${\alpha }_1$ values (b = 0.19, SE = 0.077, p = .013) and a trend-level DFA α₁ × load interaction (b = 0.12, SE = 0.067, p = .079), both indicating that blunted physiological reactivity to cognitive load correlated with steeper effort discounting. Additionally, a group × DFA ${\alpha }_1$ interaction indicated that blunted reactivity predicted steeper discounting more strongly among those with current versus remitted depression (b = −0.21, SE = 0.097, p = .034), suggesting that autonomic reactivity constrained cognitive effort, especially in current depression. This interaction was not significant with respect to healthy control subjects (b = −0.15, SE = 0.10, p = .15) (full model in Table S3).

DISCUSSION

Diminished cognitive control may partly explain why people with depression are less likely to pursue goals or more likely to ruminate. Yet, whether these deficits reflect diminished cognitive capacities or reduced motivation for cognitive effort is unclear. If people with depression exert less effort, it is furthermore unclear what constrains their motivation. Here, we showed that people with depression were less willing to exert cognitive effort, controlling for cognitive capacity. We also found evidence that a brief, sad film boosted willingness to exert effort in depression, while limited autonomic reactivity may constrain it.

The reduced willingness to exert cognitive effort is independent of abilities. Indeed, people with depression discounted more steeply than their nondepressed peers, even controlling for n-back performance. By contrast, participants with remitted depression discounted the same as control subjects, indicating that motivational deficits were specific to current depression and did not persist in remission.

Importantly, shallower discounting correlates with better global functioning in occupational, social, and psychological domains. The relationship with global functioning highlights the vital importance of motivation for cognitive effort in diverse domains of daily life. Thus, there is a need to understand underlying motivational mechanisms and how they go awry.

Aberrant striatal dopamine function is one candidate explanation of lower motivation. Robust dopamine signaling among healthy adults predicts greater sensitivity to reward benefits and lesser sensitivity to effort costs, and greater willingness to exert cognitive effort (28). Critically, people with depression have reduced reward sensitivity and striatal reward encoding (18,66–70), which may underlie lower motivation [though see (69)]. Serotonin may also play a key role. Selective serotonin reuptake inhibitors can promote physical effort for reward by selectively reducing effort cost sensitivity without altering reward sensitivity (71). Thus, diminished serotonin signaling may make people with depression more sensitive to effort costs.

We also found an intriguing, albeit counterintuitive effect of a brief, sad film, which boosted motivation to exert effort among participants with depression. One explanation is that sad films produce affective outcomes that people with depression desire, thus enhancing their motivation to pursue other goals. Perhaps people with depression prefer sad over happy media not because it increases their sadness, but because sad media is subjectively calming (72,73). Alternatively, people with depression may prefer sadness itself because it affords a form of validation about selfi-dentity (74,75). Sad media may thus boost motivation because it produces desirable affective states (e.g., calmness or validation). The effect may further depend on the locus and/or the controllability of sad content. For example, the motivational consequences of sadness induced by something bad happening in the life of a person with depression may look very different from the consequences of their sadness induced because they watch a brief sad film. While it is beyond the scope of our article, we are intrigued by the prospect that people with depression may seek out sad stimuli (films, music, etc.) [c.f., (72,73)] precisely because it amplifies a kind of sadness, which may have beneficial effects, including increased motivation. Finally, another account considers that the COGED task itself shares key features with attention refocusing strategies that people use to repair their mood (76). Perhaps attentional refocusing after a sad mood induction enhances subjective control over affect that boosts motivation for cognitive tasks. Such possibilities merit future attention, as they may elucidate a means of increasing motivation for cognitive effort in depression.

Our results extend work showing that autonomic reactivity to cognitive load is constrained in depression. DFA revealed a clear parametric effect of working memory load on heart rate dynamics and blunted reactivity specifically among participants with depression who discounted steeply. Neither PEP nor RSA reliably tracked these dimensions (Figures S2, S3), suggesting that fractal scaling was more sensitive to cognitive effort. We further note that the DFA exponent ${\alpha }_1$ correlated with individual differences in RSA, but not PEP (Figures S4, S5), suggesting that the blunted DFA profile we observed among participants with depression reflects constraints on vagal withdrawal rather than sympathetic arousal.

A blunted DFA profile is consistent with the interpretation that people with depression have lower motivation and therefore exert less effort during demanding tasks. A reverse causal account is also plausible. Namely, people with depression may have a more limited capacity to adapt, physiologically, to demanding tasks, thus constraining their motivation to exert cognitive effort. Note that people with depression show blunted autonomic reactivity to diverse stimuli, and not just cognitive demands, suggesting that their autonomic reactivity is limited in general. Importantly, the capacity for autonomic adaptability is vital (37,38): Blunted reactivity predicts future cardiovascular disease among people with depression (77). Thus, people with depression may particularly avoid contexts that tax autonomic regulatory systems, e.g., including cognitive control tasks that tap vagal function to optimize performance (78). Perhaps people with depression are more sensitive to the physiological demands of such contexts, amplifying their subjective sense of effort. An account based on peripheral autonomic capacity could explain why day-long physical effort diminishes the willingness to exert cognitive effort among healthy individuals (79). Perhaps physical effort and depression similarly deplete capacity for autonomic reactivity, making people more avoidant of additional cognitive and physical demands. Finally, we note that a tighter autonomic constraint on cognitive effort may also help explain why there was a stronger coupling between discounting and heart rate dynamics in depression than for other groups, despite the fact that discounting was related to global functioning across all groups. While many factors influence willingness to exert cognitive effort (e.g., task performance), cardiac inflexibility (77) may be more important for people with depression in that they are especially sensitive to their cardiovascular state when deciding whether to engage in a demanding task.

Key limitations of our study include the between-subjects design; future studies should address whether, within participants, sad affective media can modulate motivation for those with depression. Relatedly, we did not find any individual difference relationships between effort discounting and sadness after viewing the sad film or the degree to which the sad film increased sadness (see the Supplement), but this may in part reflect the inherently lower power of a between-subjects design and the lack of a corresponding pre-/post-discounting measure. Such effects may be detectable in within-subjects designs. Another limitation is that our design precludes testing whether heart rate fractal scaling during n-back performance varies as a function of mood, as the n-back came before the mood induction. Future work is needed to disentangle the casual mechanisms underlying the correlation between blunted autonomic reactivity, measured here by heart rate fractal scaling, and diminished motivation for cognitive effort.

Our results support the hypotheses that motivation for cognitive effort is reduced in current, but not remitted, depression, is boosted by a sad mood induction, and is higher for individuals with greater physiological capacity to respond to increasing cognitive load. Our results also call for increasing attention to heart rate fractal scaling as an index of cognitive effort that is sensitive to both objective load and motivation to do cognitive work. Finally, our results raise new questions about how sad stimuli can impact motivation in depression.