Journal of Experimental Psychology: Human Perception and Performance 2014, Vol. 40, No. 5, 2005-2021

© 2014 American Psychological Association 0096-1523/14/$ 12.00 http://dx.doi.org/10.1037/a0037522

Control of Response Timing Occurs During the Simple Reaction Time Interval but On-Line for Choice Reaction Time Dana Maslovat

Stuart T. Klapp

University of British Columbia

California State University, East Bay

Richard J. Jagacinski

Ian M. Franks

Ohio State University

University of British Columbia

The preparation of multiple element movements has been examined for decades, with no clear expla­ nation offered for the disparate results observed. Results from 2 experiments are presented and, in conjunction with previous results, a theoretical interpretation is offered regarding the preparatory processes that occur before, during and after the reaction time (RT) interval for multiple element movements during both simple and choice RT paradigms. In Experiment 1, number of elements and timing complexity were manipulated in a simple RT key-press task, using a startling acoustic stimulus to probe advance preparation. Both startle and nonstartle RT increased with number of movement elements and for a movement with increased timing complexity, providing evidence that the control of response timing occurs during the RT interval. In Experiment 2, the production of key-press movements of varying number of elements was compared in a simple versus choice RT paradigm. Results indicated that simple RT was affected by the number of elements, yet choice RT was not. Additionally, choice RT trials showed significantly longer interresponse intervals compared with those observed in simple RT trials, providing evidence for online processing in choice RT. The results of both studies, together with previous findings, suggest that planning of the timing of the onsets of the elements is prepared during simple RT, whereas planning of other aspects of the sequence of elements seems to occur in the foreperiod prior to the “go” signal. Conversely, in the choice RT paradigm, timing seems to be controlled Online. This explanation may bring closure on difficulties encountered in over 50 years of research examining response preparation for complex movements. Keywords: RT, response preparation, timing, startle, complexity

In his seminal article, The Problem o f Serial Order in Be­ havior, Lashley (1951) presented a compelling argument that

pies that cannot be explained in this way because the order in which motor elements occur changes depending on the overall action plan. Thus, aspects of the action sequence must be planned in advance of its initiation. It would seem logical that advance planning could be studied by measuring RT prior to the start of action. RT is expected to increase as a function of the complexity of the response to follow because planning would occur during RT, and a more complex response would require more time to plan. Although first examined by Freeman (1907) through the use of drawing geometric shapes of varying complexity, the first study that investigated response com­ plexity in detail was by Henry and Rogers (1960), who showed that RT increased as a function of the number of arm and hand movements involved in an action sequence. The study of RT for response sequences varying in complexity was subsequently in­ vestigated in more detail at Bell Laboratories (Sternberg, Monsell, Knoll, & Wright, 1978). These experiments involved speech and typing and, like the earlier study, demonstrated increasing RT as a function of the number of elements in the response. Although interpretation of these RT findings in terms of action planning and accessing memory related to planning appears to be straightfor­ ward, subsequent analyses and findings have demonstrated other­ wise.

behavior is typically planned in advance of its initiation. This was a radical proposal at the time because the dominant theory of serial action involved, to quote Lashley, “chains of reflexes, in which performance of each element of the series provides excitation of the next (p. 114).” Lashley argued that this re­ sponse chaining approach is not adequate by providing exam-

This article was published Online First July 28, 2014. Dana Maslovat, School of Kinesiology, University of British Columbia; Stuart T. Klapp, Department of Psychology, California State University, East Bay; Richard J. Jagacinski, Department of Psychology, The Ohio State University; Ian M. Franks, School of Kinesiology, University of British Columbia. Acknowledgments for this study go to a Natural Sciences and Engineer­ ing Research Council of Canada grant awarded to Ian M. Franks. We would also like to thank three anonymous reviewers for their constructive comments on an earlier version of this article. Correspondence concerning this article should be addressed to Dana Maslovat, School of Kinesiology, University of British Columbia, War Memorial Gymnasium 210-6081 University Boulevard, Vancouver, British Columbia, Canada, V6T 1Z1. E-mail: [email protected] 2005

2006

MASLOV AT, KLAPP, JAGACINSKI, AND FRANKS

Choice RT Versus Simple RT One complication of the RT approach concerns the distinction between choice RT and simple RT—a distinction that can be traced back to Donders (1868, 1969). In both paradigms, RT is measured from the onset of an imperative stimulus until the start of the response. In the choice RT paradigm the imperative stimulus specifies which response is required, whereas in the simple RT paradigm the required response is cued in advance and the imper­ ative stimulus is only a “go” signal. Given this distinction, the apparently straightforward RT results mentioned above (Henry & Rogers, 1960; Sternberg et al., 1978) are puzzling because they involved simple RT in which the required response is cued in advance, and thus selection of response elements and planning their sequence could have occurred prior to the start of the RT interval. If planning occurs before RT, it would seem that simple RT should be independent of the number of elements to be planned. One possible explanation for the reported simple RT results might be attributable to confounded variables such as the amount of the force required to initiate the action rather than to the complexity of the sequence. This possibility has been investigated and shown to be inadequate (for a review see Klapp, 2010, p. 109). The findings and analysis presented in the present article suggest a different solution to this puzzle regarding simple RT findings.

Application of the Notion of Hierarchy of Control to RT Analysis Considerable evidence suggests that long responses are orga­ nized as a sequence of Gestalt-like chunks of action (for a review see Klapp & Jagacinski, 2011). Thus, response planning or pro­ gramming can occur at two levels in a hierarchy—programming gestures within the individual chunks or, at a higher level, with respect to the sequence in which the chunks are to occur or the timing of the initiation of the chunks. As it turns out, the pattern of RT findings for choice RT and simple RT differ depending upon which of these levels is under investigation. At the lower level of the hierarchy, which involves within-chunk programming, the findings are straightforward. Choice RT in­ creases as a function of the complexity of a chunk of action but simple RT is independent of complexity (for a review see Klapp & Jagacinski, 2011). One example is RT prior to articulation of a single word; choice RT increases as a function of the number of syllables in the word but simple RT does not (Klapp, 2003; Klapp, Anderson, & Berrian, 1973). This pattern fits the expectation that programming must occur during choice RT because the response is not specified until the RT interval begins. In simple RT the response is specified in advance so that the response can be preprogrammed earlier rather than during RT. Thus, simple RT is independent of the complexity of single action chunk, for example, word to be articulated. The focus of the present article is on the upper level of the hierarchy in which the timing of chunk initiation is controlled. Thus, we are more interested in the how the response is initiated (including timing of the onsets of the “chunks”) as opposed to the internal details of the chunks (i.e., duration, force, etc.). The pattern of RT as a function of the number of chunks in the response is opposite from the pattern described above which applies to complexity of an individual chunk. At the upper level, simple RT

does depend on the number of chunks as initially reported by Henry and Rogers (1960) but choice RT is independent of this variable (Klapp, 1995). For example, simple RT depends on the number of words in the sequence but choice RT does not (Klapp, 2003). Strong support for the claim that the effect of number of chunks on simple RT reflects the upper level of the hierarchy of control rather than the lower level (within chunk) is the result that the slope of the relation between simple RT and number of words is independent of the number of syllables in each word (Sternberg et al., 1978). Programming of articulatory gestures for the individ­ ual words seems to be encapsulated within the lower level of the hierarchy of control and, thus, does not influence the RT with respect to the upper level. The pattern of RT results for the upper level of the hierarchy does not appear to be logical. In the simple RT paradigm, the precue should have enabled programming the response prior to RT so that simple RT should not depend on response complexity. Instead simple RT does depend on the number of chunks. In the choice RT paradigm programming would be needed after the imperative occurs because the response is not specified in advance. Thus, choice RT should increase as a function of the number of chunks. However, this increase is typically not present. The pres­ ent Experiment 1 provides the basis for understanding the first puzzle— why does simple RT depend on the number of chunks even though the sequence of chunks could have been prepro­ grammed? Experiment 2 addresses the second issue of why can choice RT be independent of the number of chunks even though this variable influences simple RT.

Three Possible Loci for Response Planning Our theoretical analysis of these issues includes specification of the types of planning processing that occur in each of three intervals: The first interval involves the time prior to the impera­ tive “go” stimulus. This applies only to the simple RT paradigm because only this paradigm provides a specification of the re­ sponse before the imperative signal. Planning that occurs during this interval can be considered advance planning, also known as preprogramming. The second interval is during the RT itself, which represents the time between the “go” signal and movement initiation. The third interval involves the time after RT, once the movement has been initiated, and is typically considered “online” programming. For example, control of some aspect of the response could occur as the action proceeds or during the interresponse intervals (IRI) between the chunks.

Evidence for Programming in the Foreperiod Prior to Simple RT Two types of studies regarding foreperiod duration prior to RT converge to support a conclusion that in the simple RT paradigm participants process aspects of the sequence of chunks during this foreperiod. In the self-select paradigm the participants are shown the task to perform and they indicate when they have prepared the movement. Thus, the participants determine the foreperiod “study time” prior to continuing with the remaining aspects of the trial. This study time increases as a function of the number of chunks in the response to be made. This result indicates that more time is used to preprogram a larger number of chunks. However, simple

RESPONSE TIMING CONTROL

RT also increases as a function of number of chunks suggesting that some processing must be postponed until the RT interval (Magnuson, Robin, & Wright, 2008). In a complementary approach the duration of the foreperiod between the cue that specifies the response and the imperative “go” signal is controlled by the experimenter (Alouche, Sant’Anna, Biagioni, & Ribeiro-do-Valle, 2012). In replication of many previous findings, simple RT prior to multiple-element movements was longer than RT for single-element movements. This difference in RT as a function of number of elements was much larger when the foreperiod was very brief (300 ms) than when it was longer (1,000 ms or 2,000 ms). These results can be interpreted by assuming that at least two distinct planning pro­ cesses are involved. If the foreperiod is too brief (300 ms) to permit any substantial planning in advance of RT, both planning processes must occur during RT, thereby producing a relationship between number of elements and simple RT. At a longer forep­ eriod (1,000 ms) some processing can occur during the foreperiod rather than during simple RT; this reduces but does not eliminate the effect of number of response elements on RT. This smaller effect of number of elements on simple RT remains relatively constant as the foreperiod is lengthened further (2,000 ms) sug­ gesting that no amount of time would be sufficient to permit all processing to occur during the foreperiod. Thus, even though some planning occurs in the foreperiod, another aspect of planning must be postponed until the RT interval. Beyond these studies regarding duration of the foreperiod, ad­ ditional evidence for planning during the foreperiod has been shown through the use of a loud, acoustic stimulus, capable of eliciting a startle response. When the “go” signal is replaced with a startling acoustic stimulus (SAS) in a simple RT task, responses are typically observed at very short latency (i.e., < 100 ms), which has been attributed to the automatic triggering of a prepared response through an atypical initiation process (Carlsen, Maslovat, & Franks, 2012; Carlsen, Maslovat, Lam, Chua, & Franks, 2011; Rothwell, 2006; Valls-Sole, Kumru, & Kofler, 2008 for reviews). This short latency response triggering does not occur when ad­ vance preparation is not possible such as during a choice RT paradigm (Carlsen, Chua, Inglis, Sanderson, & Franks, 2004; Maslovat, Hodges, Chua, & Franks, 2011), supporting the notion that responses are prepared in advance during simple RT and thus subject to response triggering. This triggering effect during a simple RT paradigm has been shown for both single and multiple element movements, implying that both individual components and their sequence are preprogrammed (Maslovat et al., 2011). A final line of evidence supporting programming during the forep­ eriod is that increasing the difficulty of the movement which will occur increases the amplitude of the evoked potentials recorded over parieto-occipital areas during the foreperiod (850 ms in duration) of a simple RT paradigm (Kourtis, Sebanz, & Knoblich,

2007

ing that this approach may be fruitful. For example, Franks and van Donkelaar (1990) found a typical sequence length effect on simple RT when the timing demands of the multiple component movement were strict, but no effect when the timing demands were lenient. In a follow-up study, van Donkelaar and Franks (1991) found that isochronous movement patterns (equal duration be­ tween responses) were performed at a faster RT as compared with nonisochronous patterns, further suggesting an explanatory role for movement timing in simple RT differences. Thus, the first challenge for the present analysis is to identify the types of processing that can occur during the foreperiod and those that can occur only during simple RT. Experiment 1 provides the basis for addressing this issue by identifying the underlying vari­ able which influences simple RT. Once the variable that influences simple RT has been identified by Experiment 1, it will be possible to address the other half of the RT puzzle— why is the role of this variable not also reflected in choice RT? This issue is addressed in Experiment 2.

Experiment 1: What Type of Processing Occurs During Simple RT? In the present study, we examined the effects of timing com­ plexity on simple RT by requiring participants to produce either a single element key-press or multiple element key-press sequences with either isochronous or nonisochronous timing structure. We also used an SAS on selected trials to trigger the prepared move­ ment and thus probe what aspects of the response were prepared in advance. We predicted that for nonstartle trials, the single element movement would have a shorter RT than the multiple element movements and that the isochronous response would have a shorter RT than the nonisochronous responses. We expected the SAS to trigger the prepared responses at short latency due to a faster initiation pathway, but we were most interested in how the multiple element movements would be performed on startle trials. If, as predicted, the timing initiation of the elements cannot be performed in advance and must be done in the RT interval, two possible outcomes could be considered. One possibility is that the multiple element responses would be triggered by the SAS at short latency; however, this automatic initiation would not allow for the preparation of accurate movement timing, thus producing move­ ment with either less accurate or more variable timing. A second alternative is that the responses triggered by the SAS would not be initiated until the timing was prepared during the RT interval. In this case, the responses would still be triggered faster than non­ startle trials (due to less time to perceive the more intense “go” stimulus), but the pattern of RT with respect to timing complexity would be similar for startle and nonstartle trials (i.e., longer RT for nonisochronous movements) due to similar processes that must occur following the “go” signal but prior to movement initiation.

2012). It therefore appears likely that although preparation of some aspects of a sequenced movement are preprogrammed, there is another process that occurs during the RT interval which is af­ fected by the number of movement chunks. We suggest that the movement components and serial ordering can be preprogrammed, but the timing of initiation of each movement element cannot be prepared in advance and is thus responsible for the effect of sequence length on simple RT. There is previous research indicat­

Method Participants. Fourteen right-handed volunteers with no obvi­ ous upper body abnormalities or sensory or motor dysfunctions participated in the study after giving informed consent. However, two participants were excluded from the analysis due to a lack of consistent activation in the sternocleidomastoid muscle during startle trials, which is considered a reliable indicator of a reflexive

2008

MASLOV AT, KLAPP, JAGACINSKI, AND FRANKS

startle response (Carlsen et al., 2011). Thus, data from 12 partic­ ipants is presented (eight male, four female; M = 25.7 years, SD = 5.3 years), who were naive to the hypothesis under investigation. The study was conducted in accordance with ethical guidelines established by the University of British Columbia and conformed to the latest revision of the Declaration of Helsinki. Apparatus. Participants sat in a height-adjustable chair in front of a 22-inch color monitor (Acer X233W, 1152 X 864 pixels, 75 Hz refresh) resting on a table. Attached to the-table in front of the participant was a telegraph key (Western Union Design, #808kl) requiring two N to close, which participants used to perform various key-press movement sequences. Participants were asked to keep their fingers straight such that opening and closing of the switch was achieved through wrist (rather than finger) extension.1 Movement onset and key-press durations and intervals were measured through the contact with the telegraph key switch, which presented a voltage of 10 V when open and 0 V when closed. Surface EMG data were collected from the superficial muscle belly of the right extensor carpi radialis (ECR, agonist) and right and left sternocleidomastoid (SCM, startle indicator) using pream­ plified surface electrodes connected via shielded cabling to an external amplifier system (Delsys Model DS-80). Recording sites were prepared and cleansed in order to decrease electrical imped­ ance. The electrodes were oriented parallel to the muscle fibers, and then attached using double sided adhesive strips. A grounding electrode was placed on the participant’s left ulnar styloid process. A customized LabView® computer program controlled stimulus and feedback presentation, and initiated data collection at a rate of 1 kHz (National Instruments, PC-MIO-16E-1) 500 ms before the presentation of the “go” signal and terminated data collection 2,500 ms following the “go” signal. EMG burst onset was defined as the point at which the EMG first began a sustained rise of two standard deviations above baseline levels (mean of 100 ms of EMG activity preceding the go signal). Task and procedures. All trials began with the word “Ready!” appearing on the screen, followed by a visual cue in the center of the monitor and an auditory template of the required movement, in which tones (80 dB, 150 ms, 500 Hz) represented the timing of the key-press closures. Participants performed one of four movement sequences which included: (a) a single key-press (visual cue = “dit”), hereafter referred to as “ 1-press;” (b) a three key-press sequence with an isochronous timing pattern of 300 ms between the end of each element and the beginning of the next element (visual cue = “dit dit dit”), hereafter referred to as “3-press ISO;” (c) a three key-press sequence with a nonisochronous short interval of 150 ms between the end of Element 1 and start of Element 2 and long interval of 450 ms between the end of Element 2 and the start of Element 32 (visual cue = “dit dit dit”), hereafter referred to as “3-press SL;” (d) a three key-press se­ quence with a nonisochronous long interval of 450 ms between the end of Element 1 and start of Element 2 and short interval of 150 ms between the end of Element 2 and the start of Element 3 (visual cue = “dit dit dit”), hereafter referred to as “3-press LS.” All key-press durations were intended to be 150 ms in duration and our manipulation of interval timing ensured all of the three key move­ ments had an equal total duration of 1,050 ms. The visual precue stayed on the monitor for the duration of the foreperiod, which was randomly selected between 3,500 ms and 4,500 ms, at which time

the “go” signal was presented, to which participants were in­ structed to react as “fast and accurately as possible.” The “go” signal could either consist of a nonstartling stimulus (80 ± 2 dB, 40 ms, 1,000 Hz) or startling stimulus (124 ± 2 dB, 40 ms, 1,000 Hz, < 1 ms rise time). All auditory signals were generated by a customized computer program and were amplified and presented via a loudspeaker placed directly behind the head of the partici­ pant. The acoustic stimulus intensities were measured using the “A” weighted impulse setting of a sound level meter (Cirrus Research model CR:252B) at a distance of 30 cm from the loud­ speaker (approximately the distance to the ears of the participant). Participants performed a single testing session, lasting approx­ imately 60 min. Each testing session began with a trial in which the SAS was unexpectedly presented, but no movement was required, to ensure a reflexive startle response was present (with those participants not exhibiting SCM activation on this trial excluded from the study). Next participants performed 10 nonstartle practice trials of each movement, beginning with 1-press, followed by 3-press ISO, then 3-press SL, and finally 3-press LS. Following practice, participants performed five blocks of testing trials, each consisting of five trials of each movement randomly presented plus one startle trial of each movement pseudorandomly presented such that there were never two consecutive startle trials or startle trials on the first two trials of each block. This resulted in 24 trials per block for a total of 120 trials: 25 nonstartle and five startle trials per movement condition. Following each trial, feedback of the participant’s performance showing the closing and opening of the telegraph key during the movement was displayed beneath the template of the movement, along with the displacement RT for that trial. A separate reward bonus of $0.05 (Canadian) was set for achieving a RT below a selected threshold (initially set to 250 ms, although adjusted for each participant based on practice block performance) as well as performing a correct movement, which required all key-press durations and intervals to be within 100 ms of the required timing. Correct performance resulted in feedback being displayed in green, and incorrect performance resulted in feedback presented in red. For both practice and testing blocks, nonstartle trials which were either too fast, (displacement RT < 80 ms), too slow (displace­ ment RT > 500 ms), or did not meet the timing criteria were considered bad trials and were repeated at the end of the block (startle trials were not repeated to minimize exposure to the SAS). A total of 202 nonstartle testing trials were repeated (out of 1,200 total, 16.8%), of which 32 resulted in a second bad trial and were not repeated. We excluded 18 startle trials (out of 240 total, 7.5%) due to a lack of detectable startle reflexive response in the SCM (see Carlsen et al., 2011 for additional details); however, all remaining startle trials were included in the analyses as we were 1 Although key-press tasks are often performed using fingers, we chose a wrist movement due to previous research showing that finger movements do not show the same response triggering effects as wrist or grasp move­ ments in response to a startling stimulus (Carlsen, Chua, Inglis, Sanderson, & Franks, 2009; Honeycutt, Kharouta, & Perreault, 2013). 2 Time intervals were chosen based on previous work examining key­ press durations and IRls (Klapp, 1995), ensuring movements required some form of timing (rather than moving as fast as possible) yet total movement time was short enough that online preparation would be dis­ couraged.

RESPONSE TIMING CONTROL

interested in how the SAS affected timing performance. Practice trials were not included in the analysis. Dependent measures and statistical analyses. Response la­ tency was examined through premotor RT (time from the stim­ ulus onset to agonist EMG onset) as this time interval is thought to best represent processing that occurs after the “go” signal and allowed for comparison to previous experiments involving an SAS. We analyzed premotor RT for all startle trials with a detectable SCM burst and all nonstartle trials that matched our timing criteria via a 4 Movement (1-press, 3-press ISO, 3-press SL, 3-press LS) X 2 Stimulus Type (nonstartle, startle) repeated measures analysis of variance (ANOVA). We were also inter­ ested in the variability of RT, as typically startle trials are performed at short latency with low variability while more complex movements would be expected to result in higher variability due to more processes that need to occur or increased noise in the system (see Carlsen et al., 2012 for a detailed model and description). As we were concerned that a single irregular trial in our low number of startle trials could result in large increases in variability relative to the relatively high number of control trials, we matched startle trials with randomly chosen nonstartle trials (five trials per movement, one per testing block), regardless of whether they matched our timing criteria. The standard deviation of premotor RT was compared via a 4 Movement (1-press, 3-press ISO, 3-press SL, 3-press LS) X 2 Stimulus Type (nonstartle, startle) repeated measures ANOVA. To examine task performance, we focused our analysis on the IRI between Key-Press 1 and 2 (IRI-12) and Key-Press 2 and 3 (IRI-23), as key-press duration was always constant and complex­ ity of the movement was manipulated through interval timing. We analyzed interval duration (DUR) and variability using the same randomly selected nonstartle trials as described above, such that we were not artificially reducing our error or variability by in­ creasing the number of trials or only selecting trials that met our timing criteria. Variability was calculated as variable error (VE; standard deviation of the scores) as this measure is typically used when participants have a specific target or timed interval to achieve. Duration and VE for IRI-12 and IRI-23 were each exam­ ined separately via a 3 Movement (3-press ISO, 3-press SL, 3-press LS) X 2 Stimulus Type (nonstartle, startle) repeated mea­ sures ANOVA. For all repeated measure ANOVAs, the Greenhouse-Geisser Epsilon factor was used to adjust the degrees of freedom for

2009

violations to sphericity. Uncorrected degrees of freedom are re­ ported, with the corrected p values and partial eta squared (-rip) values reported as a measure of effect size. The alpha level was set at .05, and significant results were examined via simple effects tests to determine the locus of the differences.

Results Summary. Mean values for all dependent measures are shown in Table 1, with RT data shown in Figure 1 and IRI data shown in Figure 2. Overall, our results indicated that although RTs were faster and less variable for startle trials, there was a movement complexity effect relating to both number of elements and timing complexity for both startle and nonstartle trials (see Figure 1). In terms of task performance, all three multiple component move­ ments were performed with a high degree of accuracy for interval timing during both startle and nonstartle trials (see Figure 2). However, as predicted, the SAS impaired timing performance as shown by decreased duration of IRI-12 for the 3-press ISO and 3-press LS movements and increased variability for the nonisochronous movements. Premotor reaction time. Analysis of premotor RT confirmed a main effect for stimulus type, F( 1, 11) = 45.72, p < .001, T)p = .81, and a main effect for movement, F(3, 33) = 25.68, p < .001, tip = .70; however, these effects were superseded by a significant Stimulus Type X Movement interaction F(3, 33) = 6.28, p = .007, T|p = .36. Post hoc analyses revealed that for nonstartle trials, the single key-press movement was significantly faster than all other movements (all ps < .001) and the 3-press SL movement was significantly slower than the other two three component move­ ments (3-press ISO, p = .003; 3-press LS, p = .050); whereas for startle trials the single key-press movement was significantly faster than the 3-press ISO (p = .012) and 3-press SL movement (p = .013) and the 3-press SL movement was significantly slower than the 3-press LS movement (p = .043; Figure 1A). To further examine the effect of timing complexity we performed a post hoc t test analysis on the nonstartle trials, comparing the RT of the 3-press ISO condition with the RT of the 3-press nonisochronous movements (by calculating a mean RT for the 3-press SL and 3-press LS combined). This analysis confirmed a significantly {p = .008) longer RT for the nonisochronous movements (M =

Table 1 Experiment 1 Results (Mean and Bracketed Standard Error) From the Simple Reaction Time (RT) Paradigm fo r Nonstartle and Startle Trials fo r Each Movement Nonstartle

Startle

Variable

1-press

3-press ISO

3-press SL

3-press LS

Premotor RT (ms) Premotor RT SD (ms) IRI-12 DUR (ms) IRI-23 DUR (ms) IRI-12 VE (ms) IRI-23 VE (ms)

142.6 (8.0) 28.4 (4.3) — — — —

174.0 (8.5) 53.5 (8.4) 312.4 (7.8) 316.7 (7.5) 34.6 (4.7) 33.2 (7.3)

190.2 (10.0) 37.4 (3.0) 189.3(8.9) 462.6 (10.4) 26.1 (2.4) 48.6 (5.1)

179.5(11.3) 49.8 (8.3) 487.1 (10.9) 186.7 (8.2) 46.3 (4.4) 21.7 (2.6)

1-press 92.8 (3.0) 12.8 (2.7) — — —

—

3-press ISO

3-press SL

3-press LS

108.0 (4.9) 23.8(6.1) 295.3 (9.9) 313.8(12.2) 34.9 (5.2) 35.1 (7.7)

113.0(6.7) 32.2 (8.4) 188.0(19.5) 455.1 (25.9) 53.0(11.8) 96.8 (14.9)

105.3 (5.7) 23.1 (5.3) 429.8 (22.0) 187.8(6.6) 62.5 (9.4) 24.6 (3.3)

Note. Data shown for premotor RT mean and standard deviation (SD) as well as mean duration (DUR) and variable error (VE) For the interval between Key-Press 1 and 2 (INT-12) and Key-Press 2 and 3 (INT-23).

MASLOV AT, KLAPP, JAGACINSKI, AND FRANKS

2010

A

* *

**

_______________________

■ l- p r e s s I

0 3 - p r e s s ISO □ 3 -p re s s SL □ 3 -p re s s LS

*

Non-Startle

Startle

Figure 1. Premotor RT (RT, panel A) and variability (panel B) for startle and nonstartle conditions for all four movement types. Single asterisks (*) denote differences between movements, and a double asterisk (**) denotes a main effect of stimulus type. Startle trials were performed at shorter latency and with lower variability, yet both startle and nonstartle trials showed a movement complexity effect.

185 ms) as compared with the isochronous movement (M = 174 ms). Secondary analysis: Two strategies in the 3-press LS movement. Although our results provide evidence that increased timing complexity can cause an increase in simple RT, the 3-press

LS movement was performed at a shorter latency as compared with the 3-press SL movement in both nonstartle and startle trials, even though they would be predicted to have a similar timing complexity structure as both are nonisochronous patterns. We speculate that because the 3-press LS movement had a longer first

RESPONSE TIMING CONTROL

600

2011

■ IRI-12 □ IRI-23

500 VI

E. "to 400

£



c — 60 .1), data were collapsed across blocks. Although the main effect for condition was not significant, F (l, 32) < 1, there was a significant Condition X Number of Key-Press interaction, F(2, 64) = 5.481, p = .007. Post hoc analysis of this interaction confirmed a replication of previous findings regarding RT as a function of number of chunks (N\ number of presses in this case) for the choice versus simple RT conditions (see Figure 3). Whereas

2015

simple RT increased systematically as a function of N, choice RT did not. Specifically, simple RT increased as a function of N, F(2, 32) = 20.055, p < .001. The slope of this relationship was 16.8 ms per chunk, which was a good fit to a linear regression, r2 = .93, By contrast, choice RT was relatively independent of N with a slope of only 4.5 ms per chunk, F(2, 32) = 3.624, p = .037; this effect was not monotonic and was not a good fit to a linear regression, r 2 = .43. The fact that the number of chunks plays a small but reduced role in determining choice RT is consistent with the conclusion that the strategy employed in this task may vary across participants as documented in the secondary analysis presented later. Interval timing. We hypothesized that a longer IRI would be indicative of online control of timing, whereas a shorter IRI would be indicative of advance programming. The interval between KeyPress 1 and 2 (IRI-12) was significantly (50 ms) longer in the choice RT paradigm than in the simple RT paradigm, F (l, 32) = 5.381, p = .025 (see Figure 4). This finding is consistent with our prediction that online programming would be more prevalent during choice RT. Details regarding IRI-12 indicate that this difference between RT paradigms is robust. This difference in IRI-12 between choice RT and simple RT paradigms was not reduced by practice. Instead the trend was that this difference became larger after the first block (with a difference of 32 ms, 62 ms, and 59 ms on Blocks 1, 2, and 3, respectively), although this interaction was nonsignificant F(2, 64) = 2.623, p = .079. The overall IRI-12 became shorter after the first block of testing, F(2, 64) = 4.925, p = .010 (block 1 M = 269 ms, block 2 M = 248 ms, block 3 M = 249). Overall IRI-12 was about 8 ms longer in the three-press responses than in the twopress responses, F (l, 32) = 8.95, p = .005. This difference, which is present for both choice RT and simple RT paradigms, may reflect some aspect of planning for a third response when one is required. However, the increased duration of IRI-12 in the choice RT paradigm over that in the simple RT paradigm was nearly the same (50 ms) in the 2 press and 3 press responses, F (l, 32) < 1, suggesting that the online control of the initiation of the third press, when required, did not occur during IRI-12. Instead, the analysis of IRI-23, which is described next, suggests that online control with respect to timing the initiation of the third press is carried out during the interval prior to the that response (i.e., during IRI-23). Whereas feedback regarding IRI-12 was provided during train­ ing, there was no feedback regarding IRI-23 at any stage of practice. Findings with respect to IRI-23 (see Figure 4) were similar to those reported with respect to IRI-12, and further sup­ ported the presence of online control in the choice RT paradigm; IRI-23 was longer in the choice RT paradigm than in the simple RT paradigm, F (l, 32) = 5.60, p = .023. As was the case for IRI-12, this difference in IRI-23 did not decrease over practice. Instead, there was a nonsignificant trend, F(2, 64) = 2.905, p = .060, for the difference in IRI-23 between these paradigms to increase after the first test block (32 ms) to 65 ms for Block 2 and 62 ms for Block 3. However, the overall mean IRI-23 became 4 Note that an SAS would have provided little new information as choice RT has been shown to be relatively immune to the response triggering effect of the SAS, likely due to a lack of advance preparation (Carlsen et al., 2004; Kumru et al., 2006; Maslovat et al., 2011).

MASLOVAT, KLAPP, JAGACINSKI, AND FRANKS

2016

Table 2 Experiment 2 Results (Mean and Bracketed Standard Error) From Simple and Choice Reaction Time (RT) Paradigms fo r Each Movement Simple RT

Choice RT

Variable

1-press

2-press

3-press

1-press

2-press

3-press

Displacement RT (ms) IRI-12 (ms) IRI-23 (ms)

297.9 (10.9)

322.5 (10.5) 225.4 (8.0)

331.5(11.8) 234.4 (8.2) 236.5 (8.4)

327.5 (9.5)

340.9(11.4) 211A (10.8)

336.4(10.6) 284.9 (10.4) 289.2 (10.4)

Note.

Data shown for displacement RT mean and as well as interresponse interval between Key-Press 1 and 2 (IRI-12) and Key-Press 2 and 3 (IRI-23).

shorter after the first block of testing, F(2, 64) = 5.97, p = .005, (Block 1 M = 278 ms, Block 2 M = 257 ms, Block 3 M = 254 ms).

Secondary analysis: Two strategies in the choice RT paradigm. Although these results indicate that participants in the choice RT condition can control responses online rather than during the RT interval, this strategy may be optional and could vary between participants. This section presents secondary find­ ings that are consistent with this possibility. Participants in the choice RT condition were divided into two groups of eight based on overall IRI-12 (the participant with the middle IRI-12 was omitted). For the eight participants with longest IRI-12 (M = 341 ms) the mean slope of choice RT as a function of N was slightly negative (—1.4 ms per chunk). This essentially null effect of N was expected if these participants control the onsets of the chunks on line during IRI, rather than during RT. By contrast, for the eight participants with shortest IRI-12 (M = 223 ms) the mean slope of RT as a function of N was substantial and positive (+10.1 ms per chunk), as would be expected if these participants establish timing during RT rather than during IRI. The difference in these slopes was significant as confirmed by an N by IRI-12-group interaction, F (l, 14) = 5.129, p = .038. Also con­ sistent with establishment of timing during the RT interval for participants with shorter IRI-12 was the nonsignificant (p = .156) trend for RT to be longer (mean of 371 ms) for participants with

short IRI-12 compared with those with longer IRI-12 (mean RT of 314 ms). Results from grouping based on IRI-23 corresponded to those reported above for grouping based on IRI-12. The classification of participants based on IRI-23 corresponded to the classification based on IRI-12 for 32 of the 34 participants. The mean slope for RT as a function of N was —1.2 ms per chunk for participants with long IRI-23 and +10.1 ms per chunk for participants with short IRI-23, yielding a significant N by IRI-23-group interaction, F (l, 14) = 4.934, p = .041. For simple RT there is no indication that strategies vary between participants as occurred for choice RT. Instead for participants with longest IRI-12 the slope of simple RT as a function of N was 21.0 ms per chunk; for those with shortest IRI-12 this slope was slightly smaller (15.6 ms per chunk). The corresponding results in relation to division of participants by IRI-23 were 22.9 ms and 13.7 ms per chunk.

Discussion The results of Experiment 1 were interpreted as indicating that the robust effect of number of movement chunks on simple RT can be attributed to planning the timing of the chunk onsets. Experi­ ment 2 addressed the issue of why this relationship between number of chunks and RT is typically absent in the choice RT

2 -p re s s [3 -p re s s |

2 -p r e s s |3 -p r e s s |

SRT

CRT IR I-1 2

Figure 3. Displacement RT for all three movement types during simple (SRT) and choice (CRT) RT paradigms. The asterisk (*) denotes a signif­ icant sequence length effect of increasing RT with number of elements for the SRT paradigm only.

3 -p re s s

3 -p r e s s ] SRT

CRT

IR I-2 3

Figure 4. IRI between Key-Press 1 and 2 (IRI-12) and between KeyPress 2 and 3 (IRI-23) for simple (SRT) and choice (CRT) RT paradigms. An asterisk (*) denotes a significant increase in IRI for the choice versus simple RT paradigm.

RESPONSE TIMING CONTROL

paradigm. The findings support the conclusion that this null find­ ing for choice RT is because the onsets of the chunks are triggered individually, online, during each IRI. These intervals become longer compared with those observed in the simple RT paradigm, thereby allowing time for this triggering to occur. Because this processing related to timing chunk onsets occurs after the RT interval, choice RT is independent of the number chunks. In addition, our secondary analysis suggests that for choice RT the use of the strategy of online control is optional. This possibility is supported further by results from one participant in the choice condition who produced both patterns of results on separate test runs. This participant was clearly an outlier with respect to mean IRI-23 in the planned analysis; the IRI-23 was only 139 ms compared with 289 across all 17 participants in the choice condi­ tion. (If these data had been omitted, the reported findings would have even more strongly supported the hypothesis of programming online in the choice condition. Thus, to be conservative, these data were included in the overall analysis reported above.) Upon com­ pletion of the trials called for in the experimental protocol, an additional block of trials was run for this participant with the feedback for IRI-12 turned back on. On this block the IRI-23 increased to 382 ms. The RT data also changed—from increasing as a function of N with a slope of 18 ms per chunk on the final tested block (short IRI) to becoming essentially the same across N (slope of 4 ms per chunk) on this additional block (long IRI). That pattern of RT and IRI data suggests that this participant shifted from planning the timing the chunk onsets during RT to control online. The possibility of different strategies may explain why, whereas the sequence length effect found in simple RT is robust, the sequence length in choice RT may (e.g., Maslovat et al., 2011) or may not (e.g., Klapp, 1995, 2003) show a similar effect of increasing movement elements. The optional online strategy was not present during simple RT as our secondary analysis suggested that all participants completed movement preparation during the RT interval. Corresponding to the analysis of Experiment 1, we need to consider the alternative interpretation that the findings of Experi­ ment 2 could be attributed to differences in the structuring of chunks. The more rapid responding in the simple RT paradigm might be due to representation of the response as a single chunk. In the choice RT paradigm the production of chunks is slower; this is consistent with a multiple-chunk representation. According to this possible difference in structuring responses across paradigms, the effect of number of action elements on RT should be smaller for simple RT than for choice RT. Instead, this effect is actually larger for simple RT. We conclude here, as in Experiment 1, that the pattern of RT cannot be explained in terms of a plausible change in the organization of the action sequence with respect to chunking.

General Discussion Overview The purpose of the current research was to examine the prepa­ ration of the structuring of the sequence of chunks comprising movements to determine what preparation processes occur prior to, during, and following the RT interval both when the required response is known (simple RT) and unknown (choice RT) prior to

2017

the imperative signal. In Experiment 1, we examined response preparation for single and multiple component movements with varying timing structures. Simple RT was influenced by the com­ plexity of the timing of the movements (see Figure 1), providing evidence that there is a response-dependent process that occurs during the RT interval. Based on the comparison between the 3-press SL and 3-press LS movement, this RT effect is more likely to occur when the first interpress interval is short to encourage participants to prepare the timing structure prior to movement initiation, rather than online. As previous research has shown an increase in response elements affects processes prior to the “go” signal (Alouche et al., 2012; Magnuson et al„ 2008), we propose that preparation of response timing occurs during simple RT, while planning of other aspects of the sequence of elements occurs during the foreperiod. In Experiment 2, response preparation of single and multiple element movements was compared between simple and choice RT. Replicating previous results, a sequence length effect was found only for the simple RT paradigm (see Figure 3). The new result is that IRIs were longer in the choice RT paradigm as compared with the simple RT paradigm. This is indicative of online preparation during choice RT, in which the movement is initiated as soon as possible with preparation of upcoming response components prepared during execution of the ongoing movement.

Control of Response Timing Occurs During Simple RT For the simple RT paradigm, we observed an increase in RT when the number of movement elements was increased (Experi­ ment 1 and 2), as well as when the timing was more complex due to nonisochronous initiation of elements (although only for one nonisochronous movement; Experiment 1). These results replicate and extend previous findings and suggest an explanation as to why simple RT depends on the number and timing of the response elements. We propose that the preparation of timing of the onsets of required movement components occurs during simple RT. Thus, increasing the number of movement chunks increases RT, as does increasing the complexity of the required response timing (see Figure 1). The assertion that timing initiation is responsible for simple RT effects would explain why three element movements with differing timing complexity showed different RT values (Experiment 1), even though they have the same number of components. Our explanation is also consistent with the observed RT differences between startle and nonstartle trials (Experiment 1). Although both types of trials exhibited a similar pattern of results with regards to RT and movement type, the magnitude of RT differences was considerably less in startle trials. Given the SAS may act to automatically trigger movements via increased neural activation levels (Carlsen et al., 2012), less time may be available to prepare the movement timing, resulting in smaller RT differences but a corresponding increase in timing variability and decrease in timing accuracy in the multiple element movements (see Figure 2). The conclusion that timing of chunk onsets is processed during the simple RT interval constrains theoretical interpretation of the control of timing because this implies that some aspect of the preparation of timing cannot be completed prior to simple RT even though the information needed to determine timing is available in

2018

MASLOVAT, KLAPP, JAGACINSKI, AND FRANKS

advance. In this section alternative approaches are considered that would fit this constraint. The first possibility is that unlike the representation of the sequencing of chunks, the generation of the representation of timing requires immediate implementation. Thus, programming of this representation must be postponed until the time to respond has arrived. To incorporate this type of postponement, we propose a new model of timing preparation that involves an oscillator with adjustable frequency. The oscillator is initially inactive, which corresponds to its frequency being equal to zero. Programming response timing involves establishing a pattern of nonzero fre­ quency that lasts for the duration of the response. Each maxima in the oscillatory pattern triggers a chunk. For an isochronous se­ quence of chunk onsets, the frequency stays constant before it returns to zero. For a nonisochronous sequence of chunk onsets, the frequency is varied. For example, for the timing conditions of Experiment 1 the frequency is varied from a lower frequency to a higher frequency for the L-S response or from a higher to a lower frequency for the S-L response. Given the active nature of the oscillator, the timing must be specified during the simple RT interval rather than beforehand. Specifying a nonconstant fre­ quency pattern involves greater detail, and therefore requires a longer simple RT. Specifying the duration of a longer sequence of chunks involves greater relative temporal precision, and also lengthens simple RT. Timing must be specified during simple RT not only at this upper level of control (related to the onsets of chunks), but also at the lower level of control (within a single chunk). Why then is a relation between timing complexity and simple RT typically ob­ served only with respect to the higher level and not for the lower level? The answer may relate to the duration of the unit being timed. At the lower level, implementation of timing might be achieved quickly because the duration of a single chunk is rela­ tively short. Furthermore, only the timing for the first chunk is specified during the RT interval, immediately before its implemen­ tation. The timing of the gestures within later chunks in the sequence can be specified individually during the response just before each is triggered so that this would not influence simple RT for an extended sequence. Thus, any effect of within-chunk timing complexity on simple RT is small and not readily detected. At the higher level, the timing specification extends over the longer interval of the entire response sequence. Establishment of temporal control over this much longer time frame would produce a strong and easily detected influence on simple RT.

Response Complexity May Affect Neural Activation Levels A different approach that can also explain the relationship between simple RT and the complexity of timing is based on a consideration of how a plan for a sequential action might be released. Given the evidence that aspects of the response are prepared in advance of simple RT, we can think of simple RT as being the time required to increase the neural activation of the associated cortical motor neurons to some “ignition point” or threshold level (i.e., Hanes & Schall, 1996). In this manner, faster RTs can be attributed to a shorter time required to increase acti­ vation, either due to lower response threshold (Maslovat et al., 2011; Nazir & Jacobs, 1991), higher initial activation levels

(Carlsen et al., 2012), or faster rate of activation increase (Car­ penter & Williams, 1995; Maslovat, Carter, Kennefick, & Carlsen, 2014). Thus, the observed RT differences due to movement com­ plexity may relate to differences in time required to reach thresh­ old level, rather than a separate process that occurs in the RT interval. If we assume advance preparation results in an initial level of neural activation, stochastic variability that is present within the nervous system, would result in a “noisy” preparatory activation level (see Faisal, Selen, & Wolpert, 2008 for a review). Move­ ments with additional and/or nonisochronous motor commands would require greater coordination between the various compo­ nents of the nervous system resulting in a higher level of noise and thus lower activation to ensure random noise does not unintention­ ally raise the activation level above threshold and accidently trigger the movement. In this manner, more complex movements would have lower mean preparatory activation level, resulting in a longer period of time to reach threshold (i.e., initiation time) and thus increased RT latency (see black lines and vertical arrows in Figure 5). During startle trials, it is thought that the increase in initiation-related activation begins earlier following the “go” sig­ nal, due to a faster neural pathway to reach the motor cortex, and accumulates at a faster rate (Carlsen et al., 2012; Maslovat et al., 2014; Maslovat et al., 2011) resulting in faster startle RTs (see gray lines and vertical arrows in Figure 5). However, it would be expected that the pattern of RT differences with respect to move­ ment complexity would be maintained, with a smaller magnitude of difference for startle trials (gray asterisk in Figure 5) as com­ pared with nonstartle trials (black asterisk in Figure 5) as observed in Experiment 1. Although we have suggested a lower level of preparatory acti­ vation may be due to inherent noise in the system which increases with movement complexity, it is possible that there are other reasons why neural activation is lower for more complex move­ ments. For example, if the participant is unable to prepare the timing initiation of the movement in advance (as previously out­ lined), this may result in a lower activation level due to incomplete preparation of the required response. Thus, although we have presented two viable descriptions of response preparation pro­ cesses during simple RT, it is possible that both may contribute to the observed RT differences based on movement complexity. It is important to note that the validity of our critical finding that the relationship between simple RT and number of chunks is related to complexity of the timing of the onsets of these chunks, does not rest upon either of the interpretations outlined above. Instead potential interpretations are presented to illustrate that the con­ straint that some aspect of processing related to timing cannot be completed prior to RT can be incorporated into a theory of the control of response timing.

Online Preparation Occurs During Choice RT During choice RT, there is no opportunity for planning prior to the “go” signal as the participant is unaware of the required response until the imperative stimulus is presented. As choice RT was not dependent upon the number of movement elements (see Figure 3), it appears that the preparation of response timing does not occur in the RT interval but rather must occur online after movement initiation. This interpretation is consistent with the new

RESPONSE TIMING CONTROL

2019

Figure 5.

Schematic representation of neural activation (hypothetical; as a proportion of activation threshold, shown as 100%) during response preparation and initiation for responses of low- and high-movement complex­ ity. Black lines prior to initiation-related increases indicate preparatory neural activation with low (thick line) and high (thin line) variability for low and high complexity movement respectively. Increases in initiation activation are shown following the “go” stimulus at which either a startle (gray lines) or nonstartle (black lines) stimulus is presented. Note that activation begins earlier and rises faster for startle trials resulting in faster RTs (vertical arrows); however, the pattern of RT differences due to movement complexity are maintained with a smaller magnitude of difference for startle trials (gray asterisk) as compared with nonstartle trials (black asterisk).

finding that the IR Is w ere longer in the choice R T paradigm than in the sim ple RT paradigm thereby perm itting planning during m ovem ent execution (see F igure 4). O ur results indicate that online p reparation during choice R T did not occu r for all participants and m ay be an optional strategy. Som e participants exhibited longer IRIs w ith a corresponding null effect o f R T as a function o f num ber o f elem ents, and other exhibited shorter IR Is w ith a corresponding increase in R T as num ber o f elem ents increased. In this m anner, it appears that it is possible to perform the choice R T task in a sim ilar m anner to sim ple RT, in w hich tim e w as taken to prepare response tim ing prior to m ovem ent execution. W hich strategy is im plem ented m ay depend on the individual or the instructions given. If m inim ization o f R T is the m ain objective, it w ould be expected that an online preparation strategy w ould be em ployed, w hereas if reducing IRI w as em phasized, it w ould be expected that preparation during choice R T w ould be perform ed in a m anner sim ilar to sim ple R T in w hich m ovem ent initiation is delayed until the tim ing structure o f the response is com pleted.

from tw o experim ents, to gether w ith p revious findings, w e identify w hich o f these aspects o f processing occur during each o f three intervals: prior to RT, during R T , or a fter R T (i.e., o nline during the response). E ven after 50 years o f study, there has been no satisfactory explanation for the pattern o f R T as a function o f the patterning o f chunks (K lapp, 2010). O ne puzzle is that even if a precue specifies the response, sim ple R T increases as a function o f the n um ber (N) o f chunks. It w ould seem to be logical that, because o f the precue, planning o f the sequence could have o c ­ curred during the foreperiod prior to R T rath er than during R T so that sim ple R T w ould b e independent o f N. T his issue is addressed in E xperim ent 1. T he second puzzling aspect o f previous R T findings is th at choice R T typically does not increase as a function o f N even though the resp o n se is n o t precued and thus the related planning could not be com pleted during the foreperiod. T his issue is addressed in E x perim ent 2. O ur findings m ay help resolve both puzzling aspects o f the previous R T findings. T he solution to both issues lies in understanding w hich aspects o f planning occur in each interval— foreperiod, RT, and online.

Conclusion

P re v io u s fin d in g s from o th e r lab o ra to rie s sh o w in g th a t fo re ­ p e rio d d u ra tio n in cre ases w ith N (M a g n u so n e t al., 20 0 8 ) su g ­

O u r analysis relates to planning applied to the units (chunks) that m ake up a response and the tim ing o f their onsets, rath er than to the organization w ithin the individual chunks. U sing new data

g e st th a t so m e a sp e c t o f p la n n in g the re sp o n se o c cu rs d u rin g the fo rep e rio d . S im p le R T also ty p ic a lly in c re a se s w ith N in d ic a t­ ing th a t a n o th e r a sp e ct o f p ro c essin g o c cu rs d u rin g R T . A new

MASLOV AT, KLAPP, JAGACINSKI, AND FRANKS

2020

finding from Experiment 1 is that simple RT can increase when complexity of the timing of chunk onsets is manipulated across actions, even if each action is composed of an identical number of elements. Simple RT also increased as a function of N\ that effect can also be attributed to complexity of timing because timing is more difficult if there are more onsets to be timed. We suggest that the structure of the sequence of chunks, including serial ordering, can be preprogrammed but the timing of chunk onsets cannot. Thus, simple RT reflects the time required to prepare response timing and initiate the movement, provided this preparation is performed in advance of movement initiation rather than online. Findings from Experiment 2 relate to the other aspect of the puzzling data pattern concerning why choice RT is typically independent of N even though simple RT does depend on N. The critical new result from Experiment 2 is that, compared with results from the simple RT paradigm, the IRIs in the choice RT paradigm are relatively long. As advance preparation cannot occur during choice RT, the longer IRI combined with the lack of effect of number of chunks suggests that planning of the sequence and its timing does not occur during the RT interval, but rather in an online manner during movement execution. In summary, the often-reported but puzzling pattern of simple and choice RT findings in relation to the structure of chunks making up a response together with previous findings regarding foreperiod duration and the new results of Experiments 1 and 2, can all be explained by the hypotheses that the primary factor influencing simple RT is the preparation of the timing of chunk onsets prior to movement initiation, whereas the control of chunk onset timing occurs online in a choice RT paradigm. Other aspects of the response can be planned during the fore­ period in the simple RT paradigm. This explanation fits an extensive corpus of data including previous findings and the new findings presented here, but it implies abandonment of the otherwise appealing notion that the serial ordering of action units is planned during simple RT.

References Alouche, S. R., Sant’Anna, G. N., Biagioni, G., & Ribeiro-do-Valle, L. E. (2012). Influence of cueing on the preparation and execution of un­ trained and trained complex motor responses. Brazilian Journal o f Medical and Biological Research, 45, 425-435. Carlsen, A. N., Chua, R., Inglis, J. T., Sanderson, D. J., & Franks, I. M. (2004). Can prepared responses be stored subcortically? Experimental Brain Research, 159, 301-309. doi:10.1007/s00221-004-1924-z Carlsen, A. N., Chua, R., Inglis, J. T., Sanderson, D. J., & Franks, I. M. (2009). Differential effects of startle on reaction time for finger and arm movements. Journal o f Neurophysiology, 101, 306-314. doi: 10.1152/jn .00878.2007 Carlsen, A. N., Maslovat, D., & Franks, I. M. (2012). Preparation for voluntary movement in healthy and clinical populations: Evidence from startle. Clinical Neurophysiology, 123, 21-33. doi:10.1016/j.clinph.2011 .04.028 Carlsen, A. N., Maslovat, D., Lam, M. Y., Chua, R., & Franks, I. M. (2011). Considerations for the use of a startling acoustic stimulus in studies of motor preparation in humans. Neuroscience & Biobehavioral Reviews, 35, 366-376. doi:10.1016/j.neubiorev.2010.04.009 Carpenter, R. H., & Williams, M. L. (1995). Neural computation of log likelihood in control of saccadic eye movements. Nature, 377, 59-62. doi: 10.1038/377059a0

Donders, F. C. (1868, 1969). Over de snelheid van psychische processen [On the speed of mental processes]. In W. G. Koster (Ed.), Attention and performance II. Acta Psychologica (Vol. 30, pp. 412-431), Amsterdam: North-Holland Publishing Company. Faisal, A. A., Selen, L. P., & Wolpert, D. M. (2008). Noise in the nervous system. Nature Reviews Neuroscience, 9, 292-303. doi: 10.1038/ nm2258 Franks, I. M., & van Donkelaar, P. (1990). The effects of demanding temporal accuracy on the programming of simple tapping sequences. Acta Psychologica, 74, 1-14. doi:10.1016/0001-6918(90)90031-A Freeman, F. N. (1907). Preliminary experiments on writing reactions. The Psychological Review, 8, 301-333. doi:!0.1037/h0093045 Hanes, D. P., & Schall, J. D. (1996). Neural control of voluntary movement initiation. Science, 274, 427-430. doi: 10.1126/science.274.5286.427 Henry, F. M., & Rogers, D. E. (1960). Increased response latency for complicated movements and a “memory drum” theory of neuromotor reaction. Research Quarterly, 31, 448-458. Honeycutt, C. F., Kharouta, M., & Perreault, E. J. (2013). Evidence for reticulospinal contributions to coordinated finger movements in humans. Journal o f Neurophysiology, 110, 1476-1483. doi: 10.1152/jn.00866

.2012 Klapp, S. T. (1995). Motor response programming during simple and choice reaction time: The role of practice. Journal o f Experimental Psychology: Human Perception and Performance, 21, 1015-1027. doi: 10.1037/0096-1523.21.5.1015 Klapp, S. T. (2003). Reaction time analysis of two types of motor prepa­ ration for speech articulation: Action as a sequence of chunks. Journal o f Motor Behavior, 35, 135-150. doi:10.1080/00222890309602129 Klapp, S. T. (2010). Comments on the classic Henry and Rogers (1960) paper on its 50th anniversary: Resolving the issue of simple versus choice reaction time. Research Quarterly fo r Exercise and Sport, 81, 108-112. doi: 10.1080/02701367.2010.10599634 Klapp, S. T., Anderson, W. G., & Berrian, R. W. (1973). Implicit speech in reading reconsidered. Journal o f Experimental Psychology, 100, 3 6 8 374. doi: 10.1037/h0035471 Klapp, S. T., & Jagacinski, R. J. (2011). Gestalt principles in the control of motor action. Psychological Bulletin, 137, 443-462. doi:10.1037/ a0022361 Kourtis, D., Sebanz, N„ & Knoblich, G. (2012). EEG correlates of Fitts’s law during preparation for action. Psychological Research, 76, 514-524. doi:10.1007/s00426-012-0418-z Kumru, H., Urra, X., Compta, Y., Castellote, J. M., Turbau, J., & VallsSole, J. (2006). Excitability of subcortical motor circuits in Go/no-go and forced choice reaction time tasks. Neuroscience Letters, 406, 66-70. doi: 10.1016/j .neulet.2006.07.012 Lashley, K. S. (1951). The problem of serial order in behavior. In L. A. Jeffress (Ed.), Cerebral mechanisms in behavior (pp. 112-136). New York, NY: Wiley. Magnuson, C. E., Robin, D. A., & Wright, D. L. (2008). Motor program­ ming when sequencing multiple elements of the same duration. Journal o f Motor Behavior, 40, 532-544. doi:10.3200/JMBR.40.6.532-544 Maslovat, D., Carlsen, A. N., Chua, R., & Franks, I. M. (2009). Response preparation changes during practice of an asynchronous bimanual move­ ment. Experimental Brain Research, 195, 383-392. doi: 10.1007/ s00221-009-1801 -x Maslovat, D., Carter, M. J., Kennefick, M., & Carlsen, A. N. (2014). Startle neural activity is additive with normal cortical initiation-related activa­ tion. Neuroscience Letters, 558, 164-168. doi:10.1016/j.neulet.2013.11 .009 Maslovat, D., Hodges, N. J., Chua, R., & Franks, I. M. (2011). Motor preparation and the effects of practice: Evidence from startle. Behavioral Neuroscience, 125, 226-240. doi:10.1037/a0022567 Nazir, T. A., & Jacobs, A. M. (1991). The effects of target discriminability and retinal eccentricity on saccade latencies: An analysis in terms of

RESPONSE TIMING CONTROL variable-criterion theory. Psychological Research, 53, 281-289. doi: 10.1007/BF00920481 Rothwell, J. C. (2006). The startle reflex, voluntary movement, and the reticulospinal tract. In G. Cruccu & M. Hallett (Eds.), Brainstem func­ tion and dysfunction (pp. 223-231). Amsterdam, The Netherlands: Elsevier, doi: 10.1016/S1567-424X(09)70071-6 Sternberg, S., Monsell, S., Knoll, R. L., & Wright, C. E. (1978). The latency and duration of rapid movement sequences: Comparisons of speech and typewriting. In G. E. Stelmach (Ed.), Information processing in motor control and learning (pp. 117-152). New York, NY: Academic Press.

2021

Valls-Sole, J., Kumru, H., & Kofler, M. (2008). Interaction between startle and voluntary reactions in humans. Experimental Brain Research, 187, 497-507. doi: 10.1007/s00221-008-1402-0 van Donkelaar, P„ & Franks, I. M. (1991). The preparation and initiation of simple rhythmical patterns. Human Movement Science, 10, 629-651. doi: 10.1016/0167-9457(91 )90020-X

Received January 13, 2014 Revision received May 21, 2014 Accepted June 6, 2014 ■

Copyright of Journal of Experimental Psychology. Human Perception & Performance is the property of American Psychological Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.