Differential relationship of recent self-reported stress and acute anxiety with divided attention performance

Abstract

There have been relatively few studies on the relationship between recent perceived environmental stress and cognitive performance, and the existing studies do not control for state anxiety during the cognitive testing. The current study addressed this need by examining recent self-reported environmental stress and divided attention performance, while controlling for state anxiety. Fifty-four university undergraduates who self-reported a wide range of perceived recent stress (10-item perceived stress scale) completed both single and dual (simultaneous auditory and visual stimuli) continuous performance tests. Partial correlation analysis showed a statistically significant positive correlation between perceived stress and the auditory omission errors from the dual condition, after controlling for state anxiety and auditory omission errors from the single condition (r = 0.41). This suggests that increased environmental stress relates to decreased divided attention performance in auditory vigilance. In contrast, an increase in state anxiety (controlling for perceived stress) was related to a decrease in auditory omission errors from the dual condition (r = − 0.37), which suggests that state anxiety may improve divided attention performance. Results suggest that further examination of the neurobiological consequences of environmental stress on divided attention and other executive functioning tasks is needed.

• Anxiety
• continuous performance test
• divided attention
• executive functioning
• perceived stress scale
• stress

Introduction

The constructs of “stress” and “anxiety” are closely related, and both are implicated in many psychiatric disorders—either together or separately (Glue et al. 1993; American Psychiatric Association 2000). Here, the term “stress” is considered to represent the response to an ongoing (or previous) stressor, whereas the term “anxiety” is considered to represent anticipatory worry about a future stressor. Complex neurobiological system changes triggered by the stress response in particular have been putatively linked to decreased cognitive performance. Studies consistently show stress-related decreases in declarative memory performance, postulated to be due to the disruptive effects of cortisol in the hippocampus (Shors et al. 2001; Conrad 2006). A meta-analytic review of sixteen studies of memory in humans found that cortisol administered before recall reduces declarative memory retrieval (Het et al. 2005). However, induced psychosocial stress may enhance memory recall when introduced post-learning (i.e. during consolidation) (Nielson et al. 2005).

In addition to hippocampal effects, recent studies examining rats have shown that exposure to either chronic (Cook and Wellman 2004; Liston et al. 2006) or acute (Brown et al. 2005) environmental stress (restraint) results in dendritic changes in the medial prefrontal cortex. Consistent with these findings, results from a brief neuropsychological battery in humans, following chronic cortisol administration, indicated a reversible deficit on tasks that rely on prefrontal functioning (e.g. spatial working memory strategy; Young et al. 1999). Another study revealed that, at the highest cortisol dose, cortisol selectively impaired performance on the task that placed relatively more demands on the prefrontal cortex, the working memory task (Lupien et al. 1996b).

There is a relative paucity of research on self-reported stress (involving cognitive appraisal) and cognitive performance. Most studies have used the related construct of “anxiety” rather than “stress.” One cross-sectional study that measured self-reported anxiety in healthy adult males found no correlation between self-reported anxiety and performance on a neuropsychological battery assessing attention, learning, memory, and motor speed (Waldstein et al. 1997). However, a longitudinal study found that an increase in state anxiety was related to improved performance on executive/prefrontal functioning (Kofman et al. 2006).

In contrast, a study that measured self-reported recent stress levels (rather than anxiety), using the perceived stress scale (PSS; Cohen et al. 1983), found that increased self-perceived stress was correlated with reduced performance on everyday tasks of selective and divided attention (Vedhara et al. 2000). Finally, studies that induced psychosocial stress reported that, at high loads, stress impaired working memory performance following the stress exposure (Oei et al. 2006; Robinson et al. 2008).

Thus, research suggests that performance on some complex tasks that place higher demands on the prefrontal cortex may be decreased due to the recent effects of stress. Consistent with this proposal, several studies have showed that administration of cortisol reduces working memory performance, which relies heavily on prefrontal functioning. However, relatively little research has examined the relationship between self-perceived recent stress and executive functioning tasks that place high demands on prefrontal functioning. The current study addresses this need by examining a wide range of self-perceived stress and performance on a traditional challenging divided attention task. Like most complex tasks, many different brain areas are necessary to perform tasks of divided attention, but these tasks have been shown to place substantial demands on the prefrontal region (Nebel et al. 2005). The current study includes measures of both recent stress and more acute anxiety to help differentiate relative contributions of each on cognitive performance. It was hypothesized that increasing levels of self-perceived recent stress would be related to increased errors on the divided attention task. It was further hypothesized that this effect size would be larger after covarying for state anxiety, which was previously shown to have the opposite effect on prefrontal tasks (Kofman et al. 2006).

Materials and methods

Participants

The research study was conducted after being reviewed and approved by the Institutional Review Board. Participants for this study were recruited from a population of undergraduate students enrolled in psychology courses. After giving informed consent, a total of 1266 students completed an online version of the 10-item PSS (Cohen and Williamson 1988). Items in the PSS measure the extent that individuals report their lives as unpredictable, uncontrollable, and overloaded over the past month (Cohen et al. 1983). The format of the PSS is a Likert Scale with five rating choices, ranging from “never (0)” to “very often (4)”. Some of this initial sample was excluded (N = 229) for having high potential for invalid responding, based on an Infrequency Scale (Jackson 1984; Calkins et al. 2004) and the 33-item Marlowe–Crowne social desirability scale (Crowne and Marlowe 1960).

From the remaining 1037 valid respondents: 73.5% were female; 69.5% of the respondents identified themselves as Caucasian; 17% as Hispanic; 9.1% as African American; 4.1% as Asian; 0.2% as American Indian; and 0.2% as Pacific Islander (mean age: 19.85 years, SD = 3.61, range 18–48 years). The PSS scores ranged from 4 to 36, with a mean of 18.82 and a standard deviation of 5.19. This initial online questionnaire served as a screening phase to create a pool of participants that had sufficient variability and range of stress levels to invite for in-person cognitive testing.

A semi-random subset of approximately 200 individuals from the online screening were invited to participate in the laboratory part of the study, with a qualitative attempt to include a wide and continuous distribution of available PSS scores from the online screening. This procedure was done to maximize both the range and variability of PSS scores (from the larger sample that completed the online measures) in the final sample invited to complete the additional laboratory-based cognitive testing.

Some of this invited subset from the online screening declined to participate in the laboratory-based part (N = 146), resulting in a total of 54 individuals who gave informed consent to complete cognitive testing, approximately 18.5 days after the online screening measures. Each gender was equally represented (27 females and 27 males). The participants ranged in age from 18 to 27 years (mean = 19.79, SD = 2.16 years). The respondents (70.4%) identified themselves as Caucasian; 16.7% as Hispanic; 9.3% as African American; and 3.7% as Asian. The paper-and-pencil version of the PSS was given at the time of cognitive testing, which produced scores ranging from 5 to 31, with a mean of 17.57 and a standard deviation of 6.64.

Measures

Visual continuous performance test (V-CPT)

The V-CPT task was created using the Vigil/W v. 1.3.0 software package (ForThought 1995). Stimuli were presented using a 22-in. NEC Multisync FP 2141^SB monitor and PC. Responses were collected using a standard keyboard. The CPT task was modeled after the A-X version (Wohlberg and Kornetsky 1973), in which a series of random single letters is presented and the participants are asked to press the spacebar after observing a target sequence of two letters. In the present study, the target consisted of the number “1” followed by the number “6”. This sequence occurred 20% of the time. Stimuli were presented in the center of the monitor at a constant cycle time of 1400 ms, with the target appearing for 50 ms followed by the blank screen appearing for 1350 ms. Each number was approximately 1.5 cm wide and 1.5 cm high.

The task began with one 30 s practice block consisting of 30 stimuli and 10 targets. The practice session had no decoys (i.e. a number other than “1” followed by “6” or “6” followed by a letter other than “1”). During the practice session, participants were given oral feedback from the experimenter regarding their accuracy. Following the practice session, the full task was administered, consisting of three consecutive blocks. Each block contained 24 targets and 12 decoys randomly occurring within 120 total stimuli. Non-target stimuli consisted of random numbers. The total task duration was approximately 7 min. A response was marked as a correct detection when the participant responded to target trials (“1” followed by “6”). Responses to non-targets were marked as a commission error while failure to respond to a target was marked as an omission error. The V-CPT was performed twice, once alone and once simultaneously with the Auditory Continuous performance test, in a divided attention trial.

Auditory continuous performance test (A-CPT)

The A-CPT task was created using the Vigil/W v. 1.3.0 software package (ForThought 1995). Stimuli were presented using a second PC with stereo speakers. The A-CPT task was modeled after the V-CPT described above, with several modifications. A series of random single letters was presented audibly (with a recorded male voice on computer), and the participant was asked to respond orally with the word “hit” when a target was heard (“K” followed by “A”). The researcher pressed a keyboard spacebar to enter the participant's oral response into the computerized accuracy monitoring software. Auditory stimuli were presented at a constant cycle time of 1400 ms by the computer (85 ms = target, 1315 ms = interstimulus interval). The A-CPT mirrored the same practice and full task protocols as described in the V-CPT section. The primary differences were: (1) the stimuli were letters instead of numbers; (2) stimuli were presented audibly instead of visually; and (3) participants indicated a response orally instead of pressing a keyboard spacebar.

Divided attention continuous performance test (DA-CPT)

The DA-CPT consisted of the simultaneous performance of both the V-CPT and the A-CPT on the two computers. Both the V-CPT and the A-CPT were the same versions of the tasks that were used when administered individually, and were administered using the same procedures as described above. The nature of the programs necessitated that they be run on two computers, and this required that the examiner tap the space bar of the second computer when the participant verbally responded “hit”. This precluded recording the participant's reaction time during the A-CPT. The DA-CPT began with the simultaneous presentations of the practice sessions from the individual V-CPT and A-CPT. Then the DA-CPT full task was administered with the simultaneous presentations of the individual V-CPT and A-CPT. The individual visual and verbal stimuli were not synchronized due to slight differences in the presentation times of the visual and auditory stimuli. The task lasted a total of 7 min.

The state trait anxiety inventory (STAI)

Only the “state” portion of the STAI (Spielberger et al. 1983) was administered. This is a 20-item questionnaire that is one of the most common measures of transient current (state) anxiety. The items in the “state” portion reflect an individual's perception of how he or she feels “right now, at the moment”, and not how he or she generally feels.

Procedure

The research appointment began with a detailed informed consent procedure with one of the investigators. Participants were then re-administered the PSS (paper version) to measure the level of perceived stress from the most recent one-month period, along with the state portion of the STAI. Immediately after the questionnaires were completed, participants were administered (in a fixed order): the visual CPT task (V-CPT), the auditory CPT task (A-CPT), and the divided attention CPT task (DA-CPT). For all CPT tasks, participants were oriented toward the computer screen by means of a chin/forehead rest positioned 18 cm from the monitor.

Statistical analysis

Descriptive statistics are shown as means ± SD for the different measures. The kurtosis values of the primary variables of interest were examined, and most of the CPT performance variables exceeded ± 2.0, which suggested that the majority of these variables had a non-parametric distribution. Therefore, all CPT performance variables were transformed with a square root function, which normalized the kurtosis values. These transformed CPT variables were used in all remaining analyses. Zero-order correlations were tested by calculating Pearson product-moment correlation coefficients, and in some cases partial correlation analysis was performed. Regression analysis was performed on some of the data to further evaluate correlation. SPSS 15 software was used for statistical analyses. p < 0.05 was considered significant.

Results

For all analyses using the PSS, the paper-and-pencil version from the time of cognitive testing was used, rather than the initial online screening version. The descriptive statistics for all measures are listed in Table I. Age was not correlated with PSS scores (r = 0.09 and p = 0.50). Similarly, there was no difference in PSS scores between genders, t(52) = 0.75, p = 0.45, or race categories, F(3,50) = 1.72, p = 0.17.

Table I.  Means, standard deviations, and ranges for PSS, STAI, and all continuous performance test outcome variables.

Measure	Mean	SD	Range
PSS	17.57	6.65	5–31
STAI	36.54	11.55	20–66
o-a-s	1.61	2.21	0–10
o-v-s	0.96	1.52	0–5
o-a-d	6.56	4.84	0–23
o-v-d	3.15	3.06	0–14
c-a-s	0.59	1.17	0–7
c-v-s	0.96	1.26	0–5
c-a-d	3.46	2.75	0–16
c-v-d	3.25	2.59	0–11

Note: o, omission errors; c, commission errors; a, auditory continuous performance test; v, visual continuous performance test; s, single task; d, dual task. The CPT variables are the raw (non-transformed) values. N = 54 subjects.

The zero-order Pearson correlations for the CPT variables are shown in Table II for descriptive purposes. The correlations revealed no statistically-significant zero-order correlation between the PSS or STAI measure with any of the CPT performance scores (from single or dual conditions).

Table II.  Pearson correlations among transformed continuous performance test variables, STAI, and PSS scores.

	PSS	STAI	o-a-s	o-v-s	o-a-d	o-v-d	c-a-s	c-v-s	c-a-d	c-v-d
PSS	–	0.76**	0.03	0.04	0.18	0.19	0.27	− 0.04	0.18	0.01
STAI		–	0.05	− 0.01	− 0.04	0.15	0.17	− 0.01	− 0.03	− 0.09
o-a-s			–	0.13	0.62**	0.39**	0.56**	− 0.18	0.51**	0.12
o-v-s				–	0.22	0.31*	0.19	0.42**	0.09	0.01
o-a-d					–	0.53**	0.34**	− 0.09	0.69*	0.23
o-v-d						–	0.30**	0.02	0.40**	0.25
c-a-s							–	0.07	0.28**	0.13
c-v-s								–	− 0.19	0.13
c-a-d									–	0.09
c-v-d										–

Note: All continuous performance test variables were transformed with a square root function prior to the correlations.

*p < 0.05 (two-tailed); **p < 0.01 (two-tailed); PSS, perceived stress scale; STAI, state trait inventory; o, omission errors; c, commission errors; a, auditory continuous performance test; v, visual continuous performance test; s, single task; d, dual task. N = 54 subjects.

Based on the a priori hypothesis, two partial correlation analyses were conducted to examine separately the relationship of omission and commission errors during the dual CPT condition with the PSS score. Each of these controlled for the variance in the STAI (state anxiety) scores and total number of respective errors from the single CPT condition. Results revealed that both omission errors, r(50) = 0.29, p = 0.04, and commission errors, r(50) = 0.28, p = 0.05, during the dual condition, showed a statistically significant relationship with PSS scores. To further explore the omission error finding, two partial correlation analyses examined the relationship between the PSS and omission errors in the two different sensory modalities involved in the task (auditory and visual). Results revealed that only the auditory omission errors from the dual condition showed a statistically significant relationship with PSS scores, after controlling for state anxiety and the omission errors from the single condition, r(50) = 0.41, p < 0.01 (see Figure 1).

Figure 1 The relationship between the PSS and number of omission errors from the auditory dual task condition. Data for all 54 subjects are shown. There is a significant correlation (p < 0.01); no significant relationship was found for visual omission errors in the dual task and the PSS (not shown).

To further explore the commission error finding, two partial correlation analyses examined the relationship between the PSS and commission errors in the two different sensory modalities. Neither of these partial correlations were statistically significant, although there was a statistical trend for increased commission errors in the auditory condition to relate to increased PSS scores, r(50) = 0.26, p = 0.07.

When examining a partial correlation between the STAI and auditory omission errors from the dual condition (partialing out the PSS and single condition omission errors), a statistically significant “negative” correlation was found, r(50) = − 0.37, p = 0.01. In additional exploratory analyses, the interaction of gender with the PSS score in predicting each of the dual CPT scores was examined, but none of these interactions approached statistical significance (all p values > 0.25).

Discussion

The use of the initial online screening phase (N = 1266) allowed for the investigation of a wide range of perceived stress in the smaller subset who completed cognitive testing. The state portion of the STAI was used in an attempt to account for each participant's acute anxiety immediately prior to cognitive testing, in order to isolate more ongoing stress thought to be reflected in the PSS score. Contrary to our first hypothesis, we did not find a zero-order correlation between the PSS score and any CPT performance index from the dual task condition. The lack of a zero-order PSS correlation is inconsistent with the findings from Vedhara colleagues (2000), which may reflect the cross-sectional design of the current study (in contrast to the longitudinal design of the earlier study), as well as differences in the attention measures.

In contrast, after controlling for the STAI scores and the single condition performance, we did find statistically significant positive correlation between the PSS and the combined omission errors from the dual task condition of the CPT. Therefore, when the variance shared between the PSS and STAI is partialed out, the relationship between the dual task combined omission error score and the PSS becomes statistically significant. This result suggests that it is more ongoing stress that relates to the decreased divided attention performance in this task, rather than acute transient anxiety. Notably, when examining the partial correlation between the STAI and dual auditory omission errors (partialing out the PSS), a statistically significant “negative” correlation was found. This finding suggests that acute anxiety may improve the divided attention abilities measured by this task, consistent with a recent report of improved executive/prefrontal functioning with increased state anxiety (on STAI) from others (Kofman et al. 2006).

To further investigate the omission error correlation, the overall omission errors from the dual task condition were analyzed separately by modality (visual and auditory), and only the auditory omission errors showed a statistically significant correlation with scores on the PSS, after controlling for the STAI scores. There is some evidence to show that performance on single auditory CPTs is generally poorer than performance on single visual CPTs (Borgaro et al. 2003), but the emphasis on the dual task condition in the current study reduces the impact of this difference on the primary analyses. However, it is possible that a greater range of omission errors on the auditory CPT dual task (0–23), relative to the visual modality (0–14) allowed for more sensitivity to correlate with the PSS.

Reaction time is often used as a measure of performance in CPT tasks and has been shown to increase when performing the tasks on days with higher self-reported stress (Sliwinski et al. 2006). However, in this protocol, each individual's reaction time was not directly measured in the auditory CPT. Instead, the participant verbally responded to a target and the researcher pressed the space bar to record the response in the computer. This was done to avoid the potential confounding influence of competing motor response selection on performance. For example, if both modalities required a slightly offset motor response, errors or reaction time delay may be introduced simply due to the competing motor responses rather than a reflection of divided attention performance.

The results of this study add to the growing body of literature that implicate the prefrontal cortex as being one of the regions deleteriously affected by recent stress, particularly as the prefrontal cortex has been shown to be a primary area involved with divided attention. However, other brain areas are involved in divided attention performance and only neuroimaging studies can directly assess the contribution of any particular brain area to decreased performance.

Consistent with our findings, other studies have reported reversible deficits on alternate tasks that substantially involve prefrontal functioning (e.g. spatial working memory strategy) in relation to elevated cortisol levels (Young et al. 1999). Further, studying effects of acute hydrocortisone administration on working and declarative memory revealed that, at the highest cortisol dose, cortisol selectively impaired performance on the working memory task (relatively more “prefrontal”; Lupien et al. 1996b). Finally, during a stressful examination period, significant decreases in performance were found on tests of selective and divided attention (purported “prefrontal tasks”), whereas performance on the word list recall task improved (Vedhara et al. 2000). The improvement in performance on the divided attention task with increasing levels of STAI appears contrary to findings where high levels of state anxiety have impaired performance on difficult tasks (Eysenck and Calvo 1992). However, this impairing of performance is not always found with divided attention (Hogan 2003), and the inverted ‘U’ performance effects of anxiety on cognitive processes is well established (Yerkes and Dodson 1908). Anxiety may have served to narrow the subjects' attention to the task at hand (e.g. screening out environmental distracters) thereby improving performance—even on the divided condition.

A number of limitations of the current study warrant consideration. The study participants were all college students, and the findings may be influenced by this younger, more educated, population. Future research may investigate whether these findings generalize to different populations, in particular working adults employed in high-stress occupations, as safety concerns may be an issue if divided attention is required for job performance. An additional area of interest would be whether greater divided attention deficits are shown by retirees from high-stress occupations. Another limitation is reliance on the self-report measure of recent stress. Self-report measures often have inherent problems with construct validity and potential for over- or underreporting distress. Although there was an attempt to control for both underreporting (MC scale) and inattention to content of items (infrequency scale), these scales cannot completely rule out the potential for these confounds on the PSS. The use of repeated cortisol measurements over an extended period of time (e.g. 1–2 months) may provide a more objective measure of biological recent stress that is, theoretically, more directly related to changes in cognitive function. Future research might also examine whether the effects found in the present study vary between subjects with different stress responses, and if the effects occur more strongly in individuals who are high corticosteroid responders, as opposed to those who mainly experience sympathetic nervous system stimulation.

A novel facet of the current study was controlling for the relationship between acute transient anxiety and recent ongoing stress. It does not appear that prior research has taken this relationship into account when examining cognitive performance. The potential confounding effects of transient anxiety on the relationship between recent self-perceived stress and cognitive performance may account for some of the past contradictory findings in the stress/anxiety literature. Future research might consider controlling for transient state anxiety in studies that examine recent stress and cognitive performance, particularly when examining performance on tasks that require substantial prefrontal cortex involvement.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.