52(2014), (2015),••–••. 397–406. Periodicals, Inc. Printed the USA. Psychophysiology, •• WileyWiley Periodicals, Inc. Printed in the in USA. C 2014 Society for Psychophysiological V Psychophysiological Research Research Copyright © 10.1111/psyp.12334 DOI: 10.1111/psyp.12334
Support for the interruption and protection hypotheses of prepulse inhibition of startle: Evidence from a modified Attention Network Test
TERRY D. BLUMENTHAL,a J. ZACHARY REYNOLDS,a and TAMARA E. SPENCEb a
Department of Psychology, Wake Forest University, Winston-Salem, North Carolina, USA Neuroscience Program, Wake Forest University Graduate School of Arts & Sciences, Winston-Salem, North Carolina, USA
b
Abstract The startle response may interrupt information processing (interruption hypothesis), and prepulse inhibition of startle (PPI) may protect that processing from interruption (protection hypothesis). These hypotheses were tested by measuring startle eyeblinks during an Attention Network Test (ANT), a combined flanker and cue reaction time (RT) task that measures the efficiency of multiple attentional networks. ANT trials with and without startle stimuli presented in the interval between the visual cue (prepulse) and target were compared. Results showed that the startle stimulus served as an alerting stimulus, speeding RT in the ANT. However, this reaction time speeding was most pronounced on trials with no startle response (100% PPI). This suggests that the alerting effect of the startle stimulus was attenuated by the startle response, and that PPI decreased the degree of this interference, in support of the interruption and protection hypotheses. Descriptors: Startle, Prepulse, PPI, Cognition, Interruption, Attention Network Test startle response. Although Graham did not state it as such, this can be seen as the inception of an interruption hypothesis (startle responding interrupts stimulus processing) and a protection hypothesis (an inhibitory system is activated by a prepulse in order to decrease that interruption). These hypotheses have been assumed to be true in hundreds of research reports in the past 40 years, but very few published reports exist in which they have been empirically tested by measuring actual processing of the prepulse, rather than just inferring changes in prepulse processing based on variations in PPI. The primary goal of the present study was to test the interruption and protection hypotheses by combining the startle paradigm with a computerized cognitive task, the Attention Network Test (ANT: Fan, McCandliss, Sommer, Raz, & Posner, 2002), to investigate whether or not the startle response interrupts the processing of a prepulse (the cue in the ANT), and whether that interruption is lessened by PPI. The protection hypothesis (based on Graham, 1975, 1992) suggests that two automatic processes are initiated when a weak stimulus is presented, one aimed at identification of the prepulse and the other functioning to inhibit potentially interrupting responses such as a startle response. The protection hypothesis is based on the assumption that the startle response would interfere with processing of the prepulse if not for the protective effect of PPI, and this interference can be stated as the interruption hypothesis. As such, “protection” cannot be assessed unless “interruption” is present, or at least possible. Although the interruption and protection hypotheses have face validity and have been repeatedly stated in the literature since first suggested by Graham, few empirical studies have actually tested these hypotheses to see whether a startle reflex does indeed affect
The startle reflex is an automatic reflex that occurs in response to sudden, intense stimuli. It is characterized by a series of physiological responses including muscle contractions in the face, neck, shoulders, and limbs (Koch, 1999). The most common measure of startle in humans is the electromyographic (EMG) activity of the orbicularis oculi, the muscle that closes the eyelids during a blink (Blumenthal et al., 2005; Landis & Hunt, 1939). This response can be significantly inhibited by the presentation of a weak stimulus 30–800 ms before the startle stimulus, a well-established phenomenon referred to as prepulse inhibition of startle (PPI) (Blumenthal 1999; Graham, 1975). Frances Graham’s Presidential Address to the Society for Psychophysiological Research (Graham, 1975) laid the foundation for decades of research using the startle response and its inhibition to investigate a variety of research questions in several areas of psychology. In the four decades since that address, startle research has contributed to our understanding of perception, development, psychopharmacology, cognition, personality, social interactions, and clinical conditions. Graham (1975, 1979, 1992) speculated that PPI might be an indication of the activation of an inhibitory mechanism that serves to reduce the interruption of prepulse processing caused by the
Portions of this material were presented at the Society for Psychophysiological Research meetings, Florence, Italy, October 3, 2013. We would like to express our most heartfelt thanks to Dr. Jin Fan for sharing his Attention Network Test program. Address correspondence to: Terry D. Blumenthal, Department of Psychology, Wake Forest University, 1834 Wake Forest Road, P.O. Box 7847 Reynolda Station, Winston-Salem, NC 27109-7487, USA. E-mail: [email protected]
397 1
398 2 cognitive processing. This would require a direct measure of prepulse processing, not just assessing startle response inhibition. The dearth of studies measuring prepulse processing is a bit surprising, given the variety of methods that could be used to measure the processing of a simple stimulus (e.g., priming, attentional blink, continuous performance tasks, rehearsal, etc.). However, the need to introduce a startle stimulus (and elicit a startle response) shortly after the prepulse complicates the evaluation of these hypotheses. Norris and Blumenthal (1996) investigated the interruption and protection hypotheses by examining the relationship between prepulse processing and PPI using a choice discrimination task. In two experiments, they presented prepulses (acoustic in one study, vibrotactile in the other) of low or high frequency, followed on some trials by an acoustic startle stimulus at 30–800 ms lead intervals. Participants were asked to report the identity of the prepulse on each trial by pressing keys corresponding to “high,” “low,” or “no” tone/vibration. Task accuracy was higher on trials in which startle responses were inhibited by the prepulse than on trials in which those responses were not inhibited, showing that task performance was positively related to PPI. This provided support for the interruption and protection hypotheses in that smaller startle responses were less able to disrupt task-related processing of the prepulse (Norris & Blumenthal, 1996). Other researchers have also tested these hypotheses, and the results of these studies have been mixed. Wynn, Dawson, and Schell (2004) found a significant relationship between backward masking and both visual and auditory PPI, with more PPI being found in conjunction with more recovery from backward masking. This may indicate that inhibiting startle increased prepulse processing, so that it would be less likely to be effectively masked by a subsequent stimulus. Perlstein, Fiorito, Simons, and Graham (1993) found that the estimated loudness of a weak prepulse was greater when that prepulse was followed by a startle stimulus than when it was not. This is evidence for loudness assimilation, the tendency for two closely presented stimuli to approach each other’s estimated loudness (the intense startle stimulus was judged to be less loud when preceded by a prepulse). However, this might also be seen as the accuracy of prepulse loudness estimation being impaired by the presentation of a startle stimulus, as would be expected under the interruption hypothesis. Other researchers have found no relationship between prepulse processing and PPI. Postma, Kumari, Hines, and Gray (2001) presented acoustic startle stimuli that were preceded on some trials by acoustic prepulses at lead intervals of 30, 60, or 120 ms. Participants were instructed to indicate whether a trial contained one stimulus (startle stimulus alone) or two (prepulse and startle stimulus). Accuracy increased with lead interval but was unaffected by the degree of PPI. This failure to find a relationship between PPI and task accuracy may have been due to a difference in cognitive load between this study and that of Norris and Blumenthal (1996): successful task performance in Postma et al. (2001) was based on prepulse detection, whereas the task used by Norris and Blumenthal required prepulse detection followed by discrimination of prepulse frequency. Elden and Flaten (2002, 2003) presented a tone followed 4 s later by a prepulse that was 10–20 ms shorter, which was itself followed by an acoustic startle stimulus on some trials, at lead intervals of 30–420 ms. Some participants had the task of determining whether or not the second tone (prepulse) was the same duration as the first (comparison stimulus). PPI was maximal at a 60-ms lead interval in both the task and no-task groups, but no relationship was found between PPI of startle and task accuracy.
T.D. T.D. Blumenthal, Blumenthal, J.Z. J.Z. Reynolds, Reynolds, and and T.E. Spence Elden and Flaten suggest that their failure to find a relationship between PPI and task accuracy may have been because of the high cognitive load of their task, which required holding an impression of the duration of a brief sound in working memory for 4 s, then detecting the next sound (prepulse), and then comparing the two. This high-load issue is paired with the fact that that cognitive load was not very different on correct versus incorrect trials. Taken together with evidence from Postma et al. (2001) and Norris and Blumenthal (1996), these studies suggest a quadratic relationship between PPI and task accuracy, with cognitive load impacting PPI when that load is moderate, but not when it is too low or too high. However, only a few studies have been reported on these questions, and it may be too early to arrive at a conclusion regarding the exact nature of the relationship between protection, interruption, and cognitive load. Other study designs have combined the startle paradigm with performance tasks and have found evidence that seems to argue against the interruption hypothesis, though these studies were not specifically designed to test that hypothesis. The StartReact effect is a phenomenon in which reaction time (RT) to a target decreases when a startle stimulus is presented simultaneously with the target stimulus during a reaction time task (Carlsen et al., 2009). To the extent that RT is a measure of stimulus processing, this StartReact effect is notable in its apparent opposition to the interruption hypothesis, which suggests that the startle response should slow responding to a target. However, research has shown that the StartReact effect is not dependent on an overt startle response, since the reaction time speeding effect is still observed when the startle response is inhibited by a prepulse (Valls-Solé, Kofler, Kumru, Castellote, & Sanegre, 2005) or eliminated through manipulation of startle stimulus rise time (Lipp, Kaplan, & Purkis, 2006; Reynolds & Day, 2007). The finding that StartReact is not dependent on an actual startle response suggests that it is closely related to another phenomenon known as the accessory stimulus effect (ASE), in which RT to a target stimulus can be decreased if the target is paired with an additional stimulus in a different modality (Hackley & Valle-Inclán, 1998, 1999; Jepma, Wagenmakers, Band, & Nieuwenhuis, 2008). Despite the differences between these two phenomena, the consistent finding is that RT is reduced when an accessory stimulus, whether it activates the startle response or not, is presented in conjunction with the target stimulus. This ASE may be based on a cross-modal summing of stimulus energy, with the converging events increasing the functional intensity of the target stimulus via energy integration across modalities (Jepma et al., 2008). In order to investigate whether the startle response disrupts prepulse processing, a task must allow for the measurement of cognitive processing and also be amenable to the addition of a startle stimulus. Norris and Blumenthal (1996) measured task accuracy but not RT, whereas the StartReact effect is based on speeding of RT (Carlsen et al., 2009). The present study combined an acoustic startle paradigm with a modified Attention Network Test (Fan et al., 2002), in which a task-relevant cue could also serve as a visual prepulse. The ANT was designed to evaluate and compare the efficiency of the alerting, orienting, and executive networks involved in attention, by combining a flanker task (Eriksen & Eriksen, 1974) with a cued RT task (Posner, 1980), as illustrated in Figure 1. Participants are asked to indicate the direction of a central target stimulus in a row of distractor (flanker) pointed brackets (>>>) by pressing one of two buttons representing “right” or “left.” Flankers can point in the direction that is the same as (congruent) or opposite (incongruent) that of the target. The row of
PPI PPI and and ANT ANT
3993
*
+ +
+ *
*
+
* +
* No cue
Center cue
Double cue
Spatial cue
>
>
Congruent
Incongruent
Figure 1. Schematic of the Attention Network Test. The top panel illustrates the four cue conditions, and the bottom panel illustrates the two congruency conditions used in the present studies. Adapted from Fan et al., 2002, 2005.
brackets can be presented either above or below a centrally located fixation cross. There are also four cue conditions, denoted by the presentation (or not) of asterisks at a fixed time before the target stimulus: one asterisk overlaying the fixation cross (center cue); two asterisks, one at each location above and below the fixation cross where the brackets may appear (double cue); one asterisk at the exact location where the target will appear (spatial cue); or no asterisk, with just a continuation of the fixation cross (no cue). For each trial, accuracy of target identification is determined, RT on correct trials is recorded, and the efficiency of each attentional component is calculated from the RT data on accurate trials (MacLeod et al., 2010). The Attention Network Test allows for the evaluation of three separate components of attention (alerting, orienting, and executive control) through a series of subtractions or orthogonal contrasts between specific conditions (Fan et al., 2002). The alerting component, which constitutes preparation for information processing and heightens awareness for stimulus reception, is calculated by comparing the average RT and accuracy for the double-cue condition with that of the no-cue condition. The no-cue condition provides no information about when or where the target stimulus will appear, whereas the double-cue condition provides only temporal information but no spatial information; therefore, attention must remain diffuse above and below the central fixation stimulus (Fan et al., 2002). The orienting component, which shifts attention toward the sensory information being received and selects the most important sensory signals for processing by the executive component, is evaluated by comparing the average RT and accuracy of the spatial-cue condition with that of the center-cue condition (MacLeod et al., 2010). The spatial-cue condition provides both temporal and spatial information about when and where the target will appear, thus causing attention to shift to that location.
Although the center cue and double cue are both considered to be alerting, since they provide temporal information but no predictive spatial information, one of each type of alerting cue is used to calculate the efficiency of the alerting and orienting networks so that the comparisons are orthogonal and no mean RT value is used in the independent calculation of more than one attentional network (McConnell & Shore, 2011). Executive control, which is involved in higher-level cognition and is responsible for goal-driven behavior, conflict resolution, planning, and decision making (Fan et al., 2009), is evaluated by comparing RT and accuracy for the congruent flanker trials with that for the incongruent flanker trials, regardless of cue location (MacLeod et al., 2010). Since incongruent flankers produce conflict, higher order cognitive processes are needed to resolve this conflict. In the present study, an acoustic startle stimulus was presented on some trials, in the interval between the onset of the cue and that of the target-flanker array. The ANT cue (asterisk) served as a visual prepulse in this study. Maximal inhibition for acoustic prepulses preceding acoustic startle stimuli occurs at stimulus onset asynchronies (i.e., lead intervals) around 120 ms (in the range between 60 ms and 240 ms), with PPI decreasing when prepulses are presented outside this range (Blumenthal, 1999; Graham & Murray, 1977). However, optimal lead intervals differ based on the sensory modalities of the prepulse and startle stimulus. For example, visual prepulses require a longer lead interval before PPI is seen (Neumann, Lipp, & Pretorius, 2004; Spence & Blumenthal, 2012), so a 250-ms lead interval (from cue onset to startle stimulus onset) was used in the present study. Because the cue must be sufficiently processed in order for ANT task performance to improve relative to that in the no-cue condition, the ANT may be a suitable tool for the evaluation of the protection and interruption hypotheses of PPI. Specifically, alerting
T.D. T.D. Blumenthal, Blumenthal, J.Z. J.Z. Reynolds, Reynolds, and and T.E. Spence
400 4 and orienting should be impaired by the presence of a startle response interrupting cue processing, and this interruption should be attenuated by the degree to which the cue inhibits that startle response. We evaluated interruption of prepulse processing by comparing task performance for different cue conditions with and without a startle stimulus. We hypothesized that the results of this study would provide evidence for the interruption hypothesis by showing that task performance worsened on trials with a startle response, attenuating the cue-dependent RT speeding and accuracy increase found in the ANT. Since the spatial-cue condition contains an extra level of information (both temporal and spatial) that could be disrupted by the startle response, we predicted that this condition would show the greatest decrement in performance. We also expected to find support for the protection hypothesis, in that RT would be inversely related to PPI. This relationship could be evaluated by calculating the correlation between the amount of PPI and RT on cue trials, but such an evaluation is complicated by the ASE (Jepma et al., 2008) mentioned above, wherein the startle stimulus would be expected to speed RT, whether the startle response was inhibited or not. It is also important to remember that the interruption hypothesis may apply not just to the processing of the prepulse, but to all information processing underway at the time of the elicitation of a startle response. That is, the startle response may interfere with processing of the prepulse, but it might also interfere with later processing of the startle stimulus itself. The protection hypothesis, being specific to the reduction of the startle response, would then predict that processing of both prepulse and startle stimulus would be attenuated less on trials in which the startle response is more strongly inhibited. Method Participants The study was approved by the Institutional Review Board at a university in the Southeastern United States. Participants were undergraduate college students who were recruited from the introductory psychology research pool and signed an informed consent form prior to the collection of any data. All participants received one credit toward their introductory psychology research option. After providing informed consent, each participant completed a health history questionnaire to screen for the presence of any exclusion criteria. Of the 38 participants who met inclusion criteria (no history of hearing loss, no psychiatric diagnosis, no use of stimulant medication in the past 2 months, no sinus congestion or head cold in the past week, no uncorrected vision problems), data from one participant were excluded from both ANT and startle data sets due to not having at least one correct response in every cue condition in the ANT. Therefore, 37 participants, 24 women and 13 men, aged 18–22 years, contributed viable data to the present study. Materials and Procedure Acoustic startle stimuli (50-ms duration, 100 dB(A), < 1 ms rise/ fall time broadband noise) were produced by Audacity software (Free Software Foundation), amplified with a Presonus HP4 amplifier, and delivered binaurally through Sennheiser PX200 headphones. Sound intensity was calibrated by presenting a 5-s broadband noise to a Quest 215 sound level meter, and a segment of this noise was used as the startle stimulus. Startle eyeblink EMG responses were quantified by recording the activity of the orbicularis oculi muscle with two electrodes
(InVivoMetric E220X Ag/AgCl, 4-mm recording diameter) that were filled with Synapse electrode cream and secured to the face below the left eye with adhesive electrode collars (E401M). Before electrode placement, the skin below the left eye and on the left temple was cleaned with a cotton swab soaked in 70% isopropyl alcohol. The first recording electrode was placed directly below the left pupil, between the top of the cheek and the lower eyelid, and the second was placed immediately lateral to and slightly higher than the first with a spacing of < 0.5 cm between outer edges of the electrodes. A third electrode was placed on the left temple to serve as a ground. The electrode wires were connected via a Biopac MEC100 extension cable to a Biopac EMG100 amplifier (5000 Gain, filters passing 1–500 Hz). This signal was sampled by a Biopac MP150 work station at 1000 Hz, and four versions of the EMG signal were recorded with AcqKnowledge 4.2 software (Biopac Systems, Inc., Goleta, CA) and saved for analysis: (1) the raw EMG; (2) filtered EMG (28–500 Hz), to exclude movement artifacts and folding frequencies; (3) a rectified (absolute value) version of the filtered signal; and (4) a smoothed version of the filtered and rectified signal, averaged with a five-sample boxcar filter. The testing and equipment rooms were individually shielded (copper mesh “Faraday cage”) and grounded, which significantly reduced electrical noise. An Attention Network Test experiment file (Fan et al., 2002) created with E-Prime stimulus presentation software (version 2.0.8.90; Psychology Software Tools, Inc., Sharpsburg, PA) was obtained from Dr. Jin Fan and modified to incorporate startle stimuli. In the modified ANT, each trial began with a 400-ms period containing only the fixation cross, followed by one of four cue conditions (no cue, center cue, double cue, spatial cue) for 250 ms (see Figure 1)1. The central fixation cross remained on throughout the entire session, appearing alone during the no-cue condition, and with the cues during the other cue conditions. At the end of the 250-ms cue condition, either a 100 dB(A) acoustic startle stimulus or a silent audio file was presented binaurally for 50 ms via headphones (see Figure 2). After an additional 150 ms of silence, the target-flanker array appeared for a maximum of 1,700 ms or until a response was made. After the response period, there was a varying intertrial interval (ITI) so that trials lasted an average of 10 s (ITI range of 8–12 s). Startle-stimulus-alone control trials were also interspersed randomly among the ANT trials, at the same ITI range (from a programming standpoint, these control trials were identical in timing to the cue and target trials, but without either cue or target). The ANT was presented on a 17-inch LCD monitor, and a chin rest was used to standardize the viewing angle and distance from the computer monitor (40 cm) and to reduce excess head movement that could interfere with EMG measurements. Participants were instructed to indicate the direction of the central pointed bracket in a line of five pointed brackets using a two-button immobilized computer mouse. Specifically, participants (all of whom were right-handed) were instructed to use the index finger for the left button and middle finger for the right button, and to respond as quickly and accurately as possible. A diagram of the four cue conditions was used as an aid to explain how to perform the task, and all participants confirmed their understanding of the instruc1. This line of inquiry began with a prior study (N = 35 participants, none of whom were included in the present report) in which the lead interval (from cue onset to startle stimulus onset) was 100 ms. PPI was minimal in that study, although the results were otherwise consistent with those of the reported study (see footnote 3).
PPI PPI and and ANT ANT
4015
+ 400 ms
* +
startle stimulus or silent audio file
+ 250 ms
+ 50 ms
150 ms
Mean trial duration = 10 s
Support for the interruption and protection hypotheses of prepulse inhibition of startle: Evidence from a modified Attention Network Test
TERRY D. BLUMENTHAL,a J. ZACHARY REYNOLDS,a and TAMARA E. SPENCEb a
Department of Psychology, Wake Forest University, Winston-Salem, North Carolina, USA Neuroscience Program, Wake Forest University Graduate School of Arts & Sciences, Winston-Salem, North Carolina, USA
b
Abstract The startle response may interrupt information processing (interruption hypothesis), and prepulse inhibition of startle (PPI) may protect that processing from interruption (protection hypothesis). These hypotheses were tested by measuring startle eyeblinks during an Attention Network Test (ANT), a combined flanker and cue reaction time (RT) task that measures the efficiency of multiple attentional networks. ANT trials with and without startle stimuli presented in the interval between the visual cue (prepulse) and target were compared. Results showed that the startle stimulus served as an alerting stimulus, speeding RT in the ANT. However, this reaction time speeding was most pronounced on trials with no startle response (100% PPI). This suggests that the alerting effect of the startle stimulus was attenuated by the startle response, and that PPI decreased the degree of this interference, in support of the interruption and protection hypotheses. Descriptors: Startle, Prepulse, PPI, Cognition, Interruption, Attention Network Test startle response. Although Graham did not state it as such, this can be seen as the inception of an interruption hypothesis (startle responding interrupts stimulus processing) and a protection hypothesis (an inhibitory system is activated by a prepulse in order to decrease that interruption). These hypotheses have been assumed to be true in hundreds of research reports in the past 40 years, but very few published reports exist in which they have been empirically tested by measuring actual processing of the prepulse, rather than just inferring changes in prepulse processing based on variations in PPI. The primary goal of the present study was to test the interruption and protection hypotheses by combining the startle paradigm with a computerized cognitive task, the Attention Network Test (ANT: Fan, McCandliss, Sommer, Raz, & Posner, 2002), to investigate whether or not the startle response interrupts the processing of a prepulse (the cue in the ANT), and whether that interruption is lessened by PPI. The protection hypothesis (based on Graham, 1975, 1992) suggests that two automatic processes are initiated when a weak stimulus is presented, one aimed at identification of the prepulse and the other functioning to inhibit potentially interrupting responses such as a startle response. The protection hypothesis is based on the assumption that the startle response would interfere with processing of the prepulse if not for the protective effect of PPI, and this interference can be stated as the interruption hypothesis. As such, “protection” cannot be assessed unless “interruption” is present, or at least possible. Although the interruption and protection hypotheses have face validity and have been repeatedly stated in the literature since first suggested by Graham, few empirical studies have actually tested these hypotheses to see whether a startle reflex does indeed affect
The startle reflex is an automatic reflex that occurs in response to sudden, intense stimuli. It is characterized by a series of physiological responses including muscle contractions in the face, neck, shoulders, and limbs (Koch, 1999). The most common measure of startle in humans is the electromyographic (EMG) activity of the orbicularis oculi, the muscle that closes the eyelids during a blink (Blumenthal et al., 2005; Landis & Hunt, 1939). This response can be significantly inhibited by the presentation of a weak stimulus 30–800 ms before the startle stimulus, a well-established phenomenon referred to as prepulse inhibition of startle (PPI) (Blumenthal 1999; Graham, 1975). Frances Graham’s Presidential Address to the Society for Psychophysiological Research (Graham, 1975) laid the foundation for decades of research using the startle response and its inhibition to investigate a variety of research questions in several areas of psychology. In the four decades since that address, startle research has contributed to our understanding of perception, development, psychopharmacology, cognition, personality, social interactions, and clinical conditions. Graham (1975, 1979, 1992) speculated that PPI might be an indication of the activation of an inhibitory mechanism that serves to reduce the interruption of prepulse processing caused by the
Portions of this material were presented at the Society for Psychophysiological Research meetings, Florence, Italy, October 3, 2013. We would like to express our most heartfelt thanks to Dr. Jin Fan for sharing his Attention Network Test program. Address correspondence to: Terry D. Blumenthal, Department of Psychology, Wake Forest University, 1834 Wake Forest Road, P.O. Box 7847 Reynolda Station, Winston-Salem, NC 27109-7487, USA. E-mail: [email protected]
397 1
398 2 cognitive processing. This would require a direct measure of prepulse processing, not just assessing startle response inhibition. The dearth of studies measuring prepulse processing is a bit surprising, given the variety of methods that could be used to measure the processing of a simple stimulus (e.g., priming, attentional blink, continuous performance tasks, rehearsal, etc.). However, the need to introduce a startle stimulus (and elicit a startle response) shortly after the prepulse complicates the evaluation of these hypotheses. Norris and Blumenthal (1996) investigated the interruption and protection hypotheses by examining the relationship between prepulse processing and PPI using a choice discrimination task. In two experiments, they presented prepulses (acoustic in one study, vibrotactile in the other) of low or high frequency, followed on some trials by an acoustic startle stimulus at 30–800 ms lead intervals. Participants were asked to report the identity of the prepulse on each trial by pressing keys corresponding to “high,” “low,” or “no” tone/vibration. Task accuracy was higher on trials in which startle responses were inhibited by the prepulse than on trials in which those responses were not inhibited, showing that task performance was positively related to PPI. This provided support for the interruption and protection hypotheses in that smaller startle responses were less able to disrupt task-related processing of the prepulse (Norris & Blumenthal, 1996). Other researchers have also tested these hypotheses, and the results of these studies have been mixed. Wynn, Dawson, and Schell (2004) found a significant relationship between backward masking and both visual and auditory PPI, with more PPI being found in conjunction with more recovery from backward masking. This may indicate that inhibiting startle increased prepulse processing, so that it would be less likely to be effectively masked by a subsequent stimulus. Perlstein, Fiorito, Simons, and Graham (1993) found that the estimated loudness of a weak prepulse was greater when that prepulse was followed by a startle stimulus than when it was not. This is evidence for loudness assimilation, the tendency for two closely presented stimuli to approach each other’s estimated loudness (the intense startle stimulus was judged to be less loud when preceded by a prepulse). However, this might also be seen as the accuracy of prepulse loudness estimation being impaired by the presentation of a startle stimulus, as would be expected under the interruption hypothesis. Other researchers have found no relationship between prepulse processing and PPI. Postma, Kumari, Hines, and Gray (2001) presented acoustic startle stimuli that were preceded on some trials by acoustic prepulses at lead intervals of 30, 60, or 120 ms. Participants were instructed to indicate whether a trial contained one stimulus (startle stimulus alone) or two (prepulse and startle stimulus). Accuracy increased with lead interval but was unaffected by the degree of PPI. This failure to find a relationship between PPI and task accuracy may have been due to a difference in cognitive load between this study and that of Norris and Blumenthal (1996): successful task performance in Postma et al. (2001) was based on prepulse detection, whereas the task used by Norris and Blumenthal required prepulse detection followed by discrimination of prepulse frequency. Elden and Flaten (2002, 2003) presented a tone followed 4 s later by a prepulse that was 10–20 ms shorter, which was itself followed by an acoustic startle stimulus on some trials, at lead intervals of 30–420 ms. Some participants had the task of determining whether or not the second tone (prepulse) was the same duration as the first (comparison stimulus). PPI was maximal at a 60-ms lead interval in both the task and no-task groups, but no relationship was found between PPI of startle and task accuracy.
T.D. T.D. Blumenthal, Blumenthal, J.Z. J.Z. Reynolds, Reynolds, and and T.E. Spence Elden and Flaten suggest that their failure to find a relationship between PPI and task accuracy may have been because of the high cognitive load of their task, which required holding an impression of the duration of a brief sound in working memory for 4 s, then detecting the next sound (prepulse), and then comparing the two. This high-load issue is paired with the fact that that cognitive load was not very different on correct versus incorrect trials. Taken together with evidence from Postma et al. (2001) and Norris and Blumenthal (1996), these studies suggest a quadratic relationship between PPI and task accuracy, with cognitive load impacting PPI when that load is moderate, but not when it is too low or too high. However, only a few studies have been reported on these questions, and it may be too early to arrive at a conclusion regarding the exact nature of the relationship between protection, interruption, and cognitive load. Other study designs have combined the startle paradigm with performance tasks and have found evidence that seems to argue against the interruption hypothesis, though these studies were not specifically designed to test that hypothesis. The StartReact effect is a phenomenon in which reaction time (RT) to a target decreases when a startle stimulus is presented simultaneously with the target stimulus during a reaction time task (Carlsen et al., 2009). To the extent that RT is a measure of stimulus processing, this StartReact effect is notable in its apparent opposition to the interruption hypothesis, which suggests that the startle response should slow responding to a target. However, research has shown that the StartReact effect is not dependent on an overt startle response, since the reaction time speeding effect is still observed when the startle response is inhibited by a prepulse (Valls-Solé, Kofler, Kumru, Castellote, & Sanegre, 2005) or eliminated through manipulation of startle stimulus rise time (Lipp, Kaplan, & Purkis, 2006; Reynolds & Day, 2007). The finding that StartReact is not dependent on an actual startle response suggests that it is closely related to another phenomenon known as the accessory stimulus effect (ASE), in which RT to a target stimulus can be decreased if the target is paired with an additional stimulus in a different modality (Hackley & Valle-Inclán, 1998, 1999; Jepma, Wagenmakers, Band, & Nieuwenhuis, 2008). Despite the differences between these two phenomena, the consistent finding is that RT is reduced when an accessory stimulus, whether it activates the startle response or not, is presented in conjunction with the target stimulus. This ASE may be based on a cross-modal summing of stimulus energy, with the converging events increasing the functional intensity of the target stimulus via energy integration across modalities (Jepma et al., 2008). In order to investigate whether the startle response disrupts prepulse processing, a task must allow for the measurement of cognitive processing and also be amenable to the addition of a startle stimulus. Norris and Blumenthal (1996) measured task accuracy but not RT, whereas the StartReact effect is based on speeding of RT (Carlsen et al., 2009). The present study combined an acoustic startle paradigm with a modified Attention Network Test (Fan et al., 2002), in which a task-relevant cue could also serve as a visual prepulse. The ANT was designed to evaluate and compare the efficiency of the alerting, orienting, and executive networks involved in attention, by combining a flanker task (Eriksen & Eriksen, 1974) with a cued RT task (Posner, 1980), as illustrated in Figure 1. Participants are asked to indicate the direction of a central target stimulus in a row of distractor (flanker) pointed brackets (>>>) by pressing one of two buttons representing “right” or “left.” Flankers can point in the direction that is the same as (congruent) or opposite (incongruent) that of the target. The row of
PPI PPI and and ANT ANT
3993
*
+ +
+ *
*
+
* +
* No cue
Center cue
Double cue
Spatial cue
>
>
Congruent
Incongruent
Figure 1. Schematic of the Attention Network Test. The top panel illustrates the four cue conditions, and the bottom panel illustrates the two congruency conditions used in the present studies. Adapted from Fan et al., 2002, 2005.
brackets can be presented either above or below a centrally located fixation cross. There are also four cue conditions, denoted by the presentation (or not) of asterisks at a fixed time before the target stimulus: one asterisk overlaying the fixation cross (center cue); two asterisks, one at each location above and below the fixation cross where the brackets may appear (double cue); one asterisk at the exact location where the target will appear (spatial cue); or no asterisk, with just a continuation of the fixation cross (no cue). For each trial, accuracy of target identification is determined, RT on correct trials is recorded, and the efficiency of each attentional component is calculated from the RT data on accurate trials (MacLeod et al., 2010). The Attention Network Test allows for the evaluation of three separate components of attention (alerting, orienting, and executive control) through a series of subtractions or orthogonal contrasts between specific conditions (Fan et al., 2002). The alerting component, which constitutes preparation for information processing and heightens awareness for stimulus reception, is calculated by comparing the average RT and accuracy for the double-cue condition with that of the no-cue condition. The no-cue condition provides no information about when or where the target stimulus will appear, whereas the double-cue condition provides only temporal information but no spatial information; therefore, attention must remain diffuse above and below the central fixation stimulus (Fan et al., 2002). The orienting component, which shifts attention toward the sensory information being received and selects the most important sensory signals for processing by the executive component, is evaluated by comparing the average RT and accuracy of the spatial-cue condition with that of the center-cue condition (MacLeod et al., 2010). The spatial-cue condition provides both temporal and spatial information about when and where the target will appear, thus causing attention to shift to that location.
Although the center cue and double cue are both considered to be alerting, since they provide temporal information but no predictive spatial information, one of each type of alerting cue is used to calculate the efficiency of the alerting and orienting networks so that the comparisons are orthogonal and no mean RT value is used in the independent calculation of more than one attentional network (McConnell & Shore, 2011). Executive control, which is involved in higher-level cognition and is responsible for goal-driven behavior, conflict resolution, planning, and decision making (Fan et al., 2009), is evaluated by comparing RT and accuracy for the congruent flanker trials with that for the incongruent flanker trials, regardless of cue location (MacLeod et al., 2010). Since incongruent flankers produce conflict, higher order cognitive processes are needed to resolve this conflict. In the present study, an acoustic startle stimulus was presented on some trials, in the interval between the onset of the cue and that of the target-flanker array. The ANT cue (asterisk) served as a visual prepulse in this study. Maximal inhibition for acoustic prepulses preceding acoustic startle stimuli occurs at stimulus onset asynchronies (i.e., lead intervals) around 120 ms (in the range between 60 ms and 240 ms), with PPI decreasing when prepulses are presented outside this range (Blumenthal, 1999; Graham & Murray, 1977). However, optimal lead intervals differ based on the sensory modalities of the prepulse and startle stimulus. For example, visual prepulses require a longer lead interval before PPI is seen (Neumann, Lipp, & Pretorius, 2004; Spence & Blumenthal, 2012), so a 250-ms lead interval (from cue onset to startle stimulus onset) was used in the present study. Because the cue must be sufficiently processed in order for ANT task performance to improve relative to that in the no-cue condition, the ANT may be a suitable tool for the evaluation of the protection and interruption hypotheses of PPI. Specifically, alerting
T.D. T.D. Blumenthal, Blumenthal, J.Z. J.Z. Reynolds, Reynolds, and and T.E. Spence
400 4 and orienting should be impaired by the presence of a startle response interrupting cue processing, and this interruption should be attenuated by the degree to which the cue inhibits that startle response. We evaluated interruption of prepulse processing by comparing task performance for different cue conditions with and without a startle stimulus. We hypothesized that the results of this study would provide evidence for the interruption hypothesis by showing that task performance worsened on trials with a startle response, attenuating the cue-dependent RT speeding and accuracy increase found in the ANT. Since the spatial-cue condition contains an extra level of information (both temporal and spatial) that could be disrupted by the startle response, we predicted that this condition would show the greatest decrement in performance. We also expected to find support for the protection hypothesis, in that RT would be inversely related to PPI. This relationship could be evaluated by calculating the correlation between the amount of PPI and RT on cue trials, but such an evaluation is complicated by the ASE (Jepma et al., 2008) mentioned above, wherein the startle stimulus would be expected to speed RT, whether the startle response was inhibited or not. It is also important to remember that the interruption hypothesis may apply not just to the processing of the prepulse, but to all information processing underway at the time of the elicitation of a startle response. That is, the startle response may interfere with processing of the prepulse, but it might also interfere with later processing of the startle stimulus itself. The protection hypothesis, being specific to the reduction of the startle response, would then predict that processing of both prepulse and startle stimulus would be attenuated less on trials in which the startle response is more strongly inhibited. Method Participants The study was approved by the Institutional Review Board at a university in the Southeastern United States. Participants were undergraduate college students who were recruited from the introductory psychology research pool and signed an informed consent form prior to the collection of any data. All participants received one credit toward their introductory psychology research option. After providing informed consent, each participant completed a health history questionnaire to screen for the presence of any exclusion criteria. Of the 38 participants who met inclusion criteria (no history of hearing loss, no psychiatric diagnosis, no use of stimulant medication in the past 2 months, no sinus congestion or head cold in the past week, no uncorrected vision problems), data from one participant were excluded from both ANT and startle data sets due to not having at least one correct response in every cue condition in the ANT. Therefore, 37 participants, 24 women and 13 men, aged 18–22 years, contributed viable data to the present study. Materials and Procedure Acoustic startle stimuli (50-ms duration, 100 dB(A), < 1 ms rise/ fall time broadband noise) were produced by Audacity software (Free Software Foundation), amplified with a Presonus HP4 amplifier, and delivered binaurally through Sennheiser PX200 headphones. Sound intensity was calibrated by presenting a 5-s broadband noise to a Quest 215 sound level meter, and a segment of this noise was used as the startle stimulus. Startle eyeblink EMG responses were quantified by recording the activity of the orbicularis oculi muscle with two electrodes
(InVivoMetric E220X Ag/AgCl, 4-mm recording diameter) that were filled with Synapse electrode cream and secured to the face below the left eye with adhesive electrode collars (E401M). Before electrode placement, the skin below the left eye and on the left temple was cleaned with a cotton swab soaked in 70% isopropyl alcohol. The first recording electrode was placed directly below the left pupil, between the top of the cheek and the lower eyelid, and the second was placed immediately lateral to and slightly higher than the first with a spacing of < 0.5 cm between outer edges of the electrodes. A third electrode was placed on the left temple to serve as a ground. The electrode wires were connected via a Biopac MEC100 extension cable to a Biopac EMG100 amplifier (5000 Gain, filters passing 1–500 Hz). This signal was sampled by a Biopac MP150 work station at 1000 Hz, and four versions of the EMG signal were recorded with AcqKnowledge 4.2 software (Biopac Systems, Inc., Goleta, CA) and saved for analysis: (1) the raw EMG; (2) filtered EMG (28–500 Hz), to exclude movement artifacts and folding frequencies; (3) a rectified (absolute value) version of the filtered signal; and (4) a smoothed version of the filtered and rectified signal, averaged with a five-sample boxcar filter. The testing and equipment rooms were individually shielded (copper mesh “Faraday cage”) and grounded, which significantly reduced electrical noise. An Attention Network Test experiment file (Fan et al., 2002) created with E-Prime stimulus presentation software (version 2.0.8.90; Psychology Software Tools, Inc., Sharpsburg, PA) was obtained from Dr. Jin Fan and modified to incorporate startle stimuli. In the modified ANT, each trial began with a 400-ms period containing only the fixation cross, followed by one of four cue conditions (no cue, center cue, double cue, spatial cue) for 250 ms (see Figure 1)1. The central fixation cross remained on throughout the entire session, appearing alone during the no-cue condition, and with the cues during the other cue conditions. At the end of the 250-ms cue condition, either a 100 dB(A) acoustic startle stimulus or a silent audio file was presented binaurally for 50 ms via headphones (see Figure 2). After an additional 150 ms of silence, the target-flanker array appeared for a maximum of 1,700 ms or until a response was made. After the response period, there was a varying intertrial interval (ITI) so that trials lasted an average of 10 s (ITI range of 8–12 s). Startle-stimulus-alone control trials were also interspersed randomly among the ANT trials, at the same ITI range (from a programming standpoint, these control trials were identical in timing to the cue and target trials, but without either cue or target). The ANT was presented on a 17-inch LCD monitor, and a chin rest was used to standardize the viewing angle and distance from the computer monitor (40 cm) and to reduce excess head movement that could interfere with EMG measurements. Participants were instructed to indicate the direction of the central pointed bracket in a line of five pointed brackets using a two-button immobilized computer mouse. Specifically, participants (all of whom were right-handed) were instructed to use the index finger for the left button and middle finger for the right button, and to respond as quickly and accurately as possible. A diagram of the four cue conditions was used as an aid to explain how to perform the task, and all participants confirmed their understanding of the instruc1. This line of inquiry began with a prior study (N = 35 participants, none of whom were included in the present report) in which the lead interval (from cue onset to startle stimulus onset) was 100 ms. PPI was minimal in that study, although the results were otherwise consistent with those of the reported study (see footnote 3).
PPI PPI and and ANT ANT
4015
+ 400 ms
* +
startle stimulus or silent audio file
+ 250 ms
+ 50 ms
150 ms
Mean trial duration = 10 s
