Acta Psychologica North-Holland
83
66 (1987) 83-102
THE PREPARATION OF ACTIONS AND PARAMETERS OF ACTION: A FIXED OR VARIABLE PROCESS? * Timothy
D. LEE, Digby ELLIOTT
and Heather
CARNAHAN
McMaster University, Hamilton, Canada Accepted
May 1987
Some information processing models consider the selection of an action plan (deciding what to do) and the parameterization of the action (deciding how to do it) as separate stages in movement preparation. Further, these models consider the stages to occur in a fixed order, whereby the selection of an action precedes the parameterization of the action. Five experiments are reported that tested this position. The first four experiments adopted a variation of Rosenbaum’s (1980) precuing technique such that, under varying conditions, different amounts and types of advance information were provided to the subject, leaving only the unspecified action or parameter to be supplied following the imperative signal. The critical finding from each study was that the time to specify an action and the time to specify a parameter were equivalent. In experiment 5, similar results were obtained using a response priming paradigm (Rosenbaum and Komblum 1982). The results of all five studies failed to support a fixed-order model of movement preparation and are more consistent with a variable-order model. The implications of these findings for other models of movement preparation are discussed.
Since the seminal work of Henry and Rogers (1960) considerable research has been conducted to assess the time course of response preparation. Reviews of this literature (e.g., Kerr 1978; Marteniuk and MacKenzie 1980) have determined that response preparation is influenced by a variety of factors directly related to the nature of the response to be produced. A partial list of these factors includes the target size to which the movement is aimed, the distance and direction of the target from the start position, and the particular limb which has * This research was funded by NSERC grants no. U0388 and no. A0406, awarded to the first and second authors, respectively. We thank John Moroz for technical assistance and Laura Diskin for preparation of the manuscript. We also thank Eric Buckolz, Christie MacKenzie and two anonymous reviewers for their comments on an earlier draft. Requests for reprints should be sent to T.D. Lee, School of Physical Education, McMaster University, Hamilton, Ontario L8S 4K1, Canada.
OOOl-6918/87/$3.50
0 1987, Elsevier Science Publishers
B.V. (North-Holland)
84
T. D. Lee et 01. / Preparing UL’~MWIS und pammeters
been determined to carry out the response. In a sense, these factors typify what we will call the paramefers of the response: each factor contributes to determining how a movement will be produced. Further, each factor has been found to contribute to the time required to prepare for movement following an imperative signal (as indexed by reaction time, RT). A neglected issue regarding response preparation chronometry, however, is the impact of what we will call the action required by the response: the specification of the goal-oriented movement plan, or the ‘what to do’ aspect of the response. Two reasons are readily apparent to explain why the nature of the action has been a neglected issue in the response preparation literature. Probably the more important of these reasons is related to the underlying purpose of the previous research: if one could determine the parameters of movement that affect response latency, then insight could be gained into the mechanisms that evolve thought into movement (or what many have called the ‘nature of the motor program’). The second reason relates to the widely held assumption that the specification of movement parameters is a late ‘stage’ in the information processing time course. Furthermore, and of particular interest in the present study, the assumption implies that parameter specification occurs after the determination of the response goal or action. Thus, by examining response factors related to how the movement was to be produced, the assumption held that parameters of the motor program could be determined independent of actual goal of the response. The assumption that response preparation is determined, to a degree, in a serial and fixed order of events is implicit in many motor programming experimental designs, although explicit theoretical statements to this end have also been made. Information processing models that specify the movement preparation process typically consider the determination of the action goal as a ‘stage’ that occurs prior to the specification of the movement parameters. As an example of this serial, fixed-order model, Theios (1973, 1975) distinguished between a response determination stage and a response program selection stage as representing a hierarchical process by which an appropriate cognitive response code must be determined before the appropriate motor program may be selected. Similarly, the motor preparation models of Requin et al. (1984) and Sanders (1980) posit that specification of the nature of a goal-directed action necessarily precedes the creation of a motor program to carry out the action. Recent theoretical propositions
T.D. Lee et al. / Preparing actions und parameters
85
regarding a more generalized motor program (e.g., Pew 1984; Schmidt 1982) also assume that the movement parameterization process occurs after a goal-oriented representation has been accessed from memory. Indeed, within an information processing framework, we know of no empirical or theoretical challenges to the assumption that the determination of what to do must precede the specification of how to do it. The goal of the five experiments reported here was to provide an initial test of what we will call the fixed-order, response preparation model. The conclusion from each of the five experiments reported here was that no support was found for the model. To our knowledge, all of the typical ‘motor programming’ research conducted since Henry and Rogers (1960) has manipulated factors that created variations in the movement kinetics and kinematics related to the response. Most of these studies were of ballistic, aiming nature, where the goal of the intended response was a simple action such as a key press or the tapping of a plate with a hand-held stylus. Regardless of whether all or some of the parameters were known in advance of the imperative signal, the action or intention of the response remained invariant: ‘depress the key’ or ‘tap the plate’ was the goal on each trial in many of these experiments. In the experiments reported here we have examined the serial, fixed-order nature of the action/parameter specification process by altering both the goal of the response as well as features related to the accomplishment of the response goal. The specification of action was determined empirically by varying the object to which the response was made. In experiments l-3 the goal of a response was either to pick up a ball or to rotate a potentiometer. In experiments 4 and 5 the goal was either to depress a telegraph key or to change the position of a toggle switch. Thus, we have empirically defined an ‘action’ here as the goal-oriented intention of a movement. As in previous research, the parameters of movement were varied through manipulation of the movement’s distance. A test of the fixed-order, response preparation model required a methodology whereby knowledge about the response action and parameters could be specified independently in advance of an imperative signal. According to the model, when only the movement parameters need to be specified following the imperative signal RT will be less than when the action must be determined following the imperative signal. Since parameters can only be specified after the determination of the action, the fixed order model holds that prior
86
T. D. Lee et al. / Preparing
actmns and pammeter.s
knowledge about the movement parameters, without prior knowledge ahout the action, will not facilitate RT relative to a condition where no prior knowledge has been provided about either the action or the parameters. In the present research two methodologies were used to test the fixed-order model. In experiments 1-4 a response precuing method was employed (cf. Goodman and Kelso 1980; Klapp 1977; Rosenbaum 1980). In experiment 5 a response priming or ‘reprogramming’ method was used (cf. Larish and Frekany 1985; Larish and Stelmach 1982; Rosenbaum and Kornblum 1982).
General method (experiments l-4) The design for experiments 1-4 was identical in many respects and will be described at this point. Variations specific to each experiment will be presented later. Subjects were tested individually in one 50-min session. Each subject participated in all conditions. The design for all four experiments was identical and consisted of three factors: (a) the type of precue condition, (b) the type of object/action signalled, and (c) the object’s distance. In all experiments there were four possible ‘target locations’ (a factorial combination of two object types and two object distances) and five precue conditions. Two precue conditions were considered as control conditions and consisted of: (1) no targets precued (a 4-choice condition), and (2) a specific target precued (a simple RT condition). The three conditions of most interest were all 2-choice conditions on which two of the four possible targets were precued in advance: (3) the two targets requiring the same action were precued (thus, distance remained uncertain, and was specified by the imperative signal), (4) the two targets requiring the same movement distance were precued (thus, action remained uncertain until the imperative signal), and (5) the two targets that required different actions and movement distances were precued (e.g., the ‘near-ball’ and ‘far-potentiometer’) (thus, while only two alternatives remained, both the required target action and distance remained unspecified). In all experiments the precues were 100% valid. Each subject performed nine trials in each of the 20 (i.e., 5 X 2 X 2) combinations of precue type, target action and target distance, for a total of 180 trials. A typical trial began with the illumination of one, two, or four of the red LEDs that were paired with each of the four targets. The LEDs illuminated at this point in the trial provided the advance information for the subject to prepare for the upcoming response. This precued information was available for 2 set and followed by either a 1.5-. 2.5 or 3.5-see delay prior to the second illumination of one of the precued LEDs (an imperative signal of I-set duration). The nine trials for each of the 20 precue combinations were divided into 3 trials at each of the three foreperiod durations. Foreperiod order was random. Instructions encouraged the subject to perform as rapidly as possible without committing errors, and to be as consistent as possible across trials in the strategies used to perform the tasks. All testing was conducted with two experimenters. The timing of
T. D. Lee et al. / Preparing actions and parameters
87
Fig. 1. Prototype of task layout. Specific target layouts (circles l-4) will be elaborated in the Method section of each experiment. The ‘X’ at the front of the apparatus represents the microswitch that the subject held down with the index finger of the preferred arm and released following the imperative signal. The four dots that appear directly in front of each target and the fouor dots that form a ‘square’ in the middle represent the sets of LEDs used in experiment 1 and
experiments 2-5, respectively.
trial events was controlled by a Lafayette four-bank timer. In experiments 1-3, RT was measured by a millisecond clock. An additional clock was added in experiment 4 to assess movement time (MT). A prototype of the apparatus that was used in each of the four
experiments
appears
in fig. 1.
The base of the apparatus was wooden, painted flat black and 45.5 cm square. The far objects were located 30 cm and the near objects were located 10 cm from the start microswitch. Each ‘outside’ LED was situated 2 cm in front of the near edge of the object. The ‘inside’ LEDs formed a 3 cm X 3 cm square (see fig. 1). The center of the square was 17 cm from the start microswitch. Median RTs (and median MTs in experiments 4 & 5) were calculated for each of the 20 precue combinations and analyzed in a 5 (precue type) x 2 (action) x 2 (distance) repeated measures ANOVA. All statistical differences are reported at the 0.05 level of significance. Post-ANOVA tests were conducted using Tukey’s HSD procedure.
Experiment 1 In the initial study we chose two tasks that were believed to require functionally distinct plans of action for execution. These tasks involved turning one of two potentiometers or grasping one of two balls. The choice of these particular objects
(tasks) was made on a purely heuristic basis. The action required if a potentiometer was signalled was a 270 o rotation of the potentiometer to the right followed by a 270° rotation to the left. The action required if a ball was signalled was to grasp and lift the ball clear of the base. Instructions encouraged the subject to perform as rapidly as possible without committing errors, and also to be as consistent as possible across trials in the strategies used to perform the tasks. Two sizes of each object were selected in order to increase the index of difficulty between the near and far target placements. The ‘large’ ball and potentiometer were located at the near target placements and the ‘small’ ball and potentiometer were situated at the ‘far’ locations. Method
Fourteen McMaster University undergraduates participated in the experiment in exchange for course credit. The sample consisted of 8 males and 6 females (M age = 21 yrs). Two of the males were left-handed. Apparatus A marble (diameter = 1.4 cm) and a squash ball (diameter = 4.1 cm) represented the small and large objects to be grasped and lifted. A small (diameter = 1.4 cm) and a large (diameter = 2.4 cm) potentiometer were the objects to be rotated. With reference to the layout represented in fig. 1, the marble was placed in target location #l, the squash ball in target location #2, the small potentiometer in target location #3 and the large potentiometer in target location #4. The ‘outside’ LEDs were used in this experiment. Procedure Trials were run in nine blocks of 20 trials. Each condition was tested once within a block of trials. The order of trials within a block was determined randomly. Results and Discussion The means for each of the 20 conditions in this experiment are illustrated in fig. 2. Of primary importance was the effect of precue type. As expected, the four-alternative (M = 357 msec) and the one-alternative (M = 263 msec) control conditions were different from the two-alternative conditions (overall M = 324 msec). Between the three. two-alternative conditions however, no differences were found. When the action type was known in advance the time to specify the distance parameter was 325 msec. When distance was known in advance, the specification of action resulted in an RT of 318 msec. When both types of information remained unspecified by the precue the RT was 327 msec. These results were confirmed statistically on the basis of a significant main effect for precue type. F(4,52) = 75.82, w2 = 0.138 and the critical value established by the Tukey test (a difference of greater than 16 msec required for statistical significance). Clearly, these findings are contrary to a fixed-order. response preparation
T. D. Lee et al. / Preparing actions and parameters
BALL
89
POTENTIOMETER
380
360
0 T C ;
340
o-o--O
1
320
a\*/
240-
-r
/
I
2 cl
1
2
2
A
DEiA
PRECUE
4
CONDITION
I
-
#
-
value
2 D
of
2 A
0
d;
0
FAR
0
NEAR 1 4 DiA
alternatIves unknown
Fig. 2. Reaction time as a function of distance (long vs. short), action (lifting a ball vs. rotating a potentiometer) and type of precue condition (# of S-R alternatives/value unknown) for experiment 1. (Note: D = distance unknown - action precued. A = action unknown - distance precued, D & A = both distance and action unknown from advance information.)
model that actions must be specified prior to their parameterization. These results suggest that the time required to specify a parameter is equivalent to the time required to specify either an action or an action and a parameter (when all three conditions have two alternatives). Two statistical interactions were also noted in the analysis. A precue type by distance interaction, F(4,52) = 2.63, o2 = 0.001, revealed that initiation latency was shorter for the near objects than the far objects in the three two-choice conditions. A precue type by action interaction, F(4,52) = 2.67, a2 = 0.002, revealed that initiation latency was shorter when a ball was required to be lifted than when a potentiometer was to be rotated under simple RT conditions. A discussion of these and other interactions to be reported in the following experiments will be presented later.
Experiment 2 The results of experiment 1 were surprising. Knowing in advance what the required action would be did not facilitate initiation latency any more than knowing what the
90
T. D. Lee et al. / Preparing uctton.~ and parumrters
distance requirements would be. This finding is contrary to a fixed-order model of movement preparation since the results suggest that an upcoming movement can be parameterized before the action is known. While the results of experiment 1 are certainly interesting, care must be exercised in the interpretation of our findings especially since variables of only marginal theoretical interest have been shown to influence the usefulness of a precue. One such variable is the spatial location of the precue and imperative signals within the task environment (Goodman and Kelso 1980). As can be seen in fig. 1, the LEDs were spatially mapped with the objects. thus providing for a potential difference in perceptual as well as movement preparation processes. In the present experiment the inside LEDs (as shown in fig. 1) were employed such that a subject was required to attend to the same central visual array location regardless of the movement preparation requested.
Method
Subjects
Twelve McMaster University undergraduates participated in exchange for course credit. The sample consisted of seven males and five females (M age = 21 yrs). Two subjects were left-handed. None of the subjects had participated in experiment 1.
Procedure
All procedures and spatial arrangements were identical to experiment 1 with one exception. In the present experiment the four objects were paired with LEDs that formed a square in the center of the apparatus (see fig. 1).
Results and Dmussion
Fig. 3 illustrates the results of the present experiment. As expected. the effect of precue type was large. F(4.44) = 23.33, o2 = 0.169. As in the previous experiment, the effect was primarily a function of the number of S-R alternatives. The four-choice condition showed the longest RT (M = 352 msec) and the simple RT condition revealed the shortest latency (M = 280 msec). Again, although the average of the three two-choice conditions fell between the two control conditions (overall M = 329 msec). there were no differences between the three precue types: action precued, distance unknown (M = 333 msec), distance precued, action unknown (M = 324 msec), and the ambiguous precue, both action and distance unknown (M = 330 msec). The critical value for a statistical difference between means was 16 msec. An interaction between precue condition, action and distance was also revealed, F(4.44) = 2.91, w2 = 0.003. Post-hoc contrasts indicated that when the far potentiometer was precued in advance, initiation latency was no faster than under the four-choice control condition. The present results replicate the findings from the previous study and argue against the possibility that the unexpected findings were due to a perceptual as opposed to
T. D. Lee et al. / Preparing actions and parameters
BALL
91
POTENTIOMETER
360 ;; E c .z
0
340
320 ? c z 0 c ” I? IL
0
\/
0 0
3oc
i 8
280
260
CYJ FAR 0
NEAR
240
i PRECUE
CONDITION
-
I
I
2 D
I
2 A
2 DELA
J
4
tt of alternatIveS value unknown
Fig. 3. Reaction time as a function of distance (long vs. short), action (lifting a ball vs. rotating a potentiometer) and type of precue condition (# of S-R alternatives/value unknown) for experiment 2. (Note: D = distance unknown - action precued, A = action unknown - distance precued, D & A = both distance and action unknown from advance information.)
motor preparation explanation. Rather, these findings point more confidently the rejection of a fixed-order, response preparation model.
towards
Experiment 3 Despite the apparent consistency of the results in the previous two studies, we had reservation about rejecting the model based upon essentially null results. In the present experiment two modifications in procedure were made in order to further explore the generalizability of our variable order response preparation findings. First, it has been suggested that trial to trial response uncertainty can influence the usefulness of advance information in this type of experimental paradigm (Goodman and Kelso 1980; Rosenbaum and Kornblum 1982). In experiments 1 and 2, a block of trials consisted of one replication of the twenty different conditions in the experiment. Within a block of trials the order of the precuing conditions was randomized. It is possible that this trial to trial variation in the type of advance information available made the movement preparation process too uncertain, thus preventing the subject from making maximum
92
T. D.
Lee et (II. / Preparing ac’trons und purumrters
use of precue information. To overcome this potential problem, the ordering of trials in the present experiment was blocked by type of precuing condition with the expectation that more effective use of the advance information could be facilitated. The second change in the present study was the spatial arrangement of the objects. In the previous experiments the near and far balls were situated to the left of the home position and the near and far potentiometers were positioned to the right. Thus, action and direction of movement were both specified by an action precue. In order to see if this specific spatial arrangement had an effect on the previous results the layout was altered.
Method
Subjects Twelve McMaster University undergraduates (M age = 20 yrs) participated in exchange for course credit. There were five male and seven female subjects. Three subjects were left-handed. No subject had participated in either experiments 1 or 2.
Apparatus In this experiment the positions of the near ball and near potentiometer were interchanged. Thus, according to fig. 1. the layout was as follows: location #1 ~ marble, location #2 - large potentiometer, location # 3 ~ small potentiometer, location #4 - squash ball. The ‘inside’ LEDs were used again.
Procedure All procedures were identical to experiment 2 with the exception of the ordering of trials. In this study the trials were run in 15 blocks of 12 trials. A block of trials consisted of three trials directed toward each of the four objects (randomly ordered). However, over the block of 12 trials, the type of precue condition remained constant. The five precue conditions were each tested once (in a random arrangement of blocks). then replicated twice (each time in a random arrangement). Prior to the start of each block of trials the subject was informed of the nature of the advance information to appear for that block of trials and encouraged to use the information to make appropriate preparation.
Results and Discussion
As in experiments 1 and 2, differences found were a function of the number of unknown S-R alternatives, but not between the three two-choice conditions, F(4.44) = 56.40. wz = 0.394. With no advance information (M = 355 msec) initiation latency was elevated relative to the two-choice conditions (overall M = 320 msec) which in turn were slower than in the simple RT situation (M = 256 msec). The three two-choice conditions resulted in mean RTs of 320, 317 and 322 msec for the distance unknown. action unknown and distance plus action unknown conditions respectively (Tukey’s critical value = 19 msec).
i? D. Lee et al. / Preparrng actions and parameters
93
POTENTIOMETER
BALL 380
360 ;; E c
340
; 320 % c 300 0’ c ::
280
: 260
0
NEAR
240 I
of
-2t PRECUE
1
I
2 D
2 A
I
A
4 DtA
CllternOtlYeS
CONDITION -
value
unknown
Fig. 4. Reaction time as a function of distance (long vs. short), action (lifting a ball vs. rotating a potentiometer) and type of precue condition (# of S-R alternatives/value unknown) for experiment 3. (Note: D = distance unknown - action precued, A = action unknown distance precued, D & A = both distance and action unknown from advance information.)
One further issue about the above finding should also be noted. The variance accounted for by the precue condition main effect ( w2 = 0.394) is more than double the variance accounted for by this manipulation in either of the two previous experiments. Apparently, while the procedural changes in the present study did not alter the conclusion of the previous two experiments, the changes did have a considerable impact on the systematic variation across the five precue conditions. The type of action as well as an action by distance interaction were also observed, F(l,ll) = 10.36, w2 = 0.001 and F(l,ll) = 17.38, w2 = 0.011, respectively. The interaction revealed that initiated latency was shorter when directed toward the near ball (M = 305 msec) than the far ball (M = 320 msec). No differences in RT were observed when comparing the near and far potentiometers (MS = 318 and 312 msec). In addition. a significant interaction between the type of precue condition and the distance was also found, F(4,44) = 3.39, w2 = 0.016. The locus of this interaction was a shorter latency when reaching for the near object than reaching for the far object under the distance unknown and action plus distance unknown, two-choice conditions.
94
Summary
T. D. Lee et crl. / Preparing actwns and parametrr.~
of experiments
l-3
To this point three studies have consistently shown that prior information regarding the action to be performed is no more beneficial than prior information about the distance to be moved - provided that two alternatives remain. In each experiment the actual differences between the means of the three two-choice conditions have been small ( -c 12 msec). Furthermore, in each study the shortest RT of these three conditions has been the action unknown conditions, which, according to a fixed response preparation model, should be slower than the condition where distance is unknown. The precue condition has interacted with type of action and distance in various, inconsistent ways over the three studies. Two issues bear against a position that these interactions contradict our primary conclusion. First, in none of the experiments did an interaction result in even one significant difference between the three two-choice precue conditions. Second, in all cases the interactions accounted for only a very small proportion of the total variance ( < 0.01). Thus, the impact and replicability of these interactions is questionable.
Experiment
4
In the present experiment we attempted to assess two critical questions that emerged from the three previous experiments. First, our results may have been particular to the objects selected and the different actions required. Our rationale in experiments 1-3 was that reaching to pick an object from the table requires a plan of action that is distinct from reaching to rotate a potentiometer. If our results correctly reject a fixed-order response preparation model then they should be replicable with two new objects, requiring supposedly distinct action plans. Our second concern was that differences between the two-choice conditions could have occurred following the initiation of movement (i.e., during the execution phase). This possibility was especially of concern because the hand configurations for grasping a ball and a potentiometer were similar. The potential existed that movement preparations prior to movement onset were related to the potential size of the object to be grasped (i.e., near/large vs. far/small) - leaving the specific decision as to the uction to be performed or the distance to traverse until during the movement. To assess the above two possibilities we replicated the design of experiment 3 with the two modifications that different objects were used and that MT was also measured.
T.D.Leeetal./ Preparing actions and parameters
95
Method Subjects Twelve McMaster University undergraduates (M age = 21 yrs) who had not participated in any of the previous studies served as subjects in exchange for course credit. The sample was comprised of five males and seven females. All of the subjects were right-handed. Apparatus The two balls and two potentiometers from the previous experiments were replaced with two telegraph keys and two toggle switches. Both telegraph keys were of identical dimensions (finger press diameter = 1.7 cm). Similarly, both toggle switches were identical (height = 1.5 cm; diameter = 0.3 cm). The telegraph keys were placed at locations 1 and 2 (in fig. 1) and the toggle switches at locations 3 and 4. Procedure All procedures in the present study were identical to the procedures of experiment 3. When a telegraph key was the object associated with the imperative signal, subjects pressed down the key with the index finger of the preferred hand. The toggle switch was a three-position-type. Prior to movement, the switch was positioned vertically (i.e., in the center position). The subject’s task was to grasp the switch between the thumb and index finger, and to move the switch to the position toward the start button. The MT clock was initiated when the subject lifted off the start button and terminated when the telegraph key was depressed or when the toggle switch left the vertical position. Results and Discussion Reuction time The type of precue condition produced results that replicated the three previous studies. The difference in RT as a function of the number of S-R alternatives accounted for the bulk of the significant main effect, F(4,44) = 86.32, w2 = 0.506. The four-choice mean RT was 362 msec, the overall two-choice mean RT was 313 msec and the mean simple RT was 246 msec. The differences between the three two-choice conditions again were small and not significant (Tukey’s critical value = 18 msec). When action was precued (distance unknown) the mean RT was 311 msec. When distance was precued (action unknown) the mean RT was 310 msec. When both action and distance were unknown in the two-choice situation the mean RT was 317 msec. These findings then, replicate the previous three studies using a new pair of objects/actions. Two further effects were also significant. There was a main effect for distance, F&11) = 17.70, w2 = 0.005, as well as an interaction between distance and action, F(l,ll) = 10.67, LJ’ = 0.001. This interaction revealed that the initiation latency prior to movement toward the near telegraph key (M = 304 msec) was less than the latency for the far key (M = 319 msec). However, the toggle switch RTs were identical for the near and far locations (both MS = 307 msec). All data points are illustrated in fig. 5.
T. D. Lee et al. / Preparing actions and parameters
96
Movement time An issue of primary concern in the present study was the effect of the precuing conditions on MT. Indeed, this main effect was small but significant, F(4,44) = 4.77, w2 = 0.001. Essentially, this effect was similar to the RT effect: the four-choice mean MT was 288 msec, the overall mean MT for the two-choice conditions was 272 msec and the MT for the simple RT condition was 262 msec. The Tukey critical value was 18 msec. Between the three two-choice conditions there were small differences: action precued, distance unknown (M = 274 msec), distance precued, action unknown (M = 268 msec) and both action and distance unknown (M = 273 msec). Interestingly. only the two-choice condition where distance was precued (action unknown) was found to differ, statistically, from the four-choice condition. From a fixed-order hierarchical model, if any carry-over effects from the RT into the MT period had been predicted, the benefit would have been in favour of the action precued (distance unknown) condition. Once again, these findings fail to support fixed-order models of movement preparation.
KEY
TOGGLE ;3
360
-
t
0 0
“E 340E
%
,0-O
320-
&I
\,/’
F 300f5 F ::
280-
:
1 i
260.
240.
f
I
2 D
2 A
PRECUE
2 DBA
Q
I
4
CONDITION
-#
FAR
.
NEAR
3
I
_
0
2 D
of Value
I
2 A
2 DBA
4
alternatIves unknown
Fig. 5. Reaction time as a function of distance (long vs. short), action (pressing a telegraph key vs. moving a toggle switch) and type of precue condition (# of S-R alternatives/value unknown) for experiment 4. (Note: D = distance unknown - action precued, A = action unknown - distance precued. D & A = both distance and action unknown from advance information.)
T.D. Lee et al. / Preparing uctions and parameters
TOGGLE
KEY
2 E
390
-
350
-
o,o-o_o~
/’
.F -
310.
2 F
270-
,/“J’ -
+ $
230
-
190
-
97
0’
/”
o--@--c
z ,” :
*-O---O
150 0’
0
FAR
l
NEAR
I I
I
4
?-
D
PRECUE
:
I
2 D
D:A
CONDITION
-
# value
of
2 A
4 DiA
alternatives unknown
Fig. 6. Movement time as a function of distance (long vs. short), action (pressing a telegraph key vs. moving a toggle switch) and type of precue condition (# of S-R alternatives/value unknown) for experiment 4. (Note: D = distance unknown - action precued, A = action unspecified distance precued, D & A = both distance and action unknown from advance information.)
All of the cell means for MT are presented in fig. 6. As expected, MT was faster for the telegraph key than for toggle switch, F(1,ll) = 83.15, a2 = 0.477. Also, short movements were made faster than long movements, F(l,ll) = 35.82, GJ’= 0.243. However, an interaction between action and distance, F(l,ll) = 14.22, o2 = 0.006 as well as a three-way interaction was found, F(444) = 3.11, w2 = 0.002. Tukey’s critical value for the three-way interaction was 19 msec. As may be seen in fig. 6, there are a few pairwise differences that signify the nature of this interaction, none of which run counter to the primary significance of these MT data.
Reaction time-movement time relationship In order to examine the relationship between RT and MT and how that relationship might vary as a function of experimental condition, within-subject correlations of RT and MT were computed from the 9 trials in each of the 20 cells. After a normalizing transformation to Fisher z-scores, these data were analyzed in a 5 (precue type) x 2 (action) x 2 (distance) repeated measures ANOVA. This analysis revealed that RT-MT relationships did not differ as a result of any of our experimental manipulations (all ps > 0.15). Since the grand mean was very close to zero (z = 0.044) we conclude that subjects were not trading-off RT and MT in this experiment.
98
T. D. Lee et al. / Preparing actions and
purmwtm
Experiment 5 Experiments 1-4, in which we used a variant of the movement precuing technique. provided no support for a serial, fixed-order model regarding the preparation of actions and parameters of action. The possibility exists however, that the precuing technique permits subjects the opportunity to devise strategies that permit complete preparation of multiple alternative responses (Rosenbaum and Kornblum 1982). For instance, in the various two-choice conditions that were critical to the tests of the serial action/parameter specification assumption in experiments l-4. it is possible that subjects fully prepared both alternative responses in advance of the imperative signal. Given such a strategy, the two-choice RT values would have assessed the time required to select one of the two fully prepared responses rather than reflecting the time required to specify the information that remained uncertain prior to the imperative signal. As an alternative to the movement precuing technique, Rosenbaum and Kornblum (1982) described a response priming technique that provides an alternative method for assessing the time course for response preparation (see also Larish and Frekany 1985; Larish and Stelmach 1982). They argued that the priming technique discourages the strategy of multiple response preparation by creating a strong subjective bias that one particular response (i.e., the primed response) will be indicated by the imperative signal. Given that only one response is prepared in advance, how is RT affected when an alternative response is indicated by the imperative signal? Using Rosenbaum and Kornblum’s method, one can investigate the preparation (or repreparation) of responses as a function of their relationship to the primed response. In the present experiment we investigated how responses are reprepared as a function of their relationship to the prepared response. The serial. fixed-order model provides a clear prediction that reprepared times will be relatively brief if the new response has the same action as the prepared response (thus, only a new parameter needs to be changed in the already prepared response). Conversely, if the new response requires either a different action or a different action and parameter, then a relatively long RT will result. However, if the priming method provides results similar to the findings from the previous four experiments then no differences would be expected as a function of the relationship between the primed and required responses. Method Subjects
Twelve undergraduate and graduate student volunteers who had not participated previously comprised the present sample. There were eight males and four females. all of whom were right-handed. The sample mean age was 25 years. Apparatus
The apparatus
was identical
to that used in experiment
4.
Procedures
The present experiment was the required response
adopted a priming method whereby the primed response on 75% of the trials. Thus, the valid/invalid trial by trial
99
T D. Lee et al. / Preparing actions and parameters Table 1 Mean median response.
RTs as a function
Relation
Required
response
Same
of target
between primed
and the relation
and required
between
the primed
response
Diff.
Diff.
Diff. action
distance
action
& distance
Far key Near key Far switch Near switch
278 265 277 273
379 365 381 377
397 376 381 375
398 385 366 411
Mean
213
376
382
390
and required
Mean
363 348 351 359
ratio was 3 : 1. Testing was conducted in five blocks of 48 trials, for a total of 240 trials per subject. Within a block of trials each of the four responses were primed twelve times. On nine of the 12 trials the primed response was valid. The three invalid primes for each response in each block of trials represented one required response for each of the alternative, non primed responses. Thus, over the course of the experiment, each response was validly primed for 45 trials. Further, each of the three non primed responses was tested five times. For each trial a response was primed by illuminating one of the four LEDs for 1 second, followed by a constant 2.5-set delay prior to the imperative signal (which was also illuminated for 1 set). While subjects were instructed that some of the primed responses would be invalid, they were encouraged to prepare to carry out the primed response since these would be valid on most of the trials. As in previous experiments, subjects were instructed to respond as fast as possible but not to make errors. Median RT and MT values were computed for each of the four targets in each of the four conditions: (a) primed and required responses being the same, (b) required response involving a different distance than what was primed, (c) required response involving a different action, or (d) required response involving a different action and distance. Statistical analyses were conducted using a 4 (primed/required response condition) X 2 (action) X 2 (distance) repeated measures model. The RT and MT means, similar to Rosenbaum and Komblum’s presentation, appear in tables 1 and 2. Results and Discussion Reaction
time
The primary result of the RT analysis was a main effect for the relationship between the primed and required response, F(3,33) = 53.16, w2 = 0.34. As a necessary requirement for the validity of the priming technique (Rosenbaum and Kornblum 1982) the RT for the valid priming condition was less (M = 273 msec) than the RT for the three invalid priming conditions (overall M = 383 msec). There were no differences between these three invalid conditions, however. When the required response involved a change in distance from the response that had been primed the mean RT was 376 msec. When
T. D. Lee et al. / Preparing
100 Table 2 Mean median response.
MTs as a function
Required response
Relation
of target
between primed
act~om and parameters
and the relation
and required
between
the primed
response
Same
Diff. distance
Diff. action
Diff. actlon & distance
Far key Near key Far switch Near switch
256 140 352 247
302 162 394 287
217 163 370 297
217 151 389 298
Mean
249
286
277
279
and
required
Mean
278 154 316 282
needed to be reprepared the mean RT was 382 msec. When both the action and distance were different from the primed response the mean RT was 390 msec. The Tukey’s critical value required for significance between means was 29 msec. These findings support the RT results of the previous four experiments and, collectively, provide no evidence that parameters of movement are specified only after an action has been selected. The only other significant RT effect in the present study was an interaction between revealed that the action and distance, F(1.11) = 5.66, L? = 0.01. This interaction latency prior to moving toward the far telegraph key (M = 363 msec) was longer than the latency for the near key (M = 348). The RTs for the far and near toggle switches were not different (MS = 351 and 359, respectively). This result replicates the RT interaction found for these tasks in experiment 4. the action
Movement
time
The effect of repreparation also made a significant contribution to the MT results, F(3.33) = 5.64, w* = 0.02. This effect paralleled the RT effect as the valid condition resulted in faster MTs (M = 249 msec) than the three invalid conditions (overall M = 281 msec), which, themselves, did not differ. Again, these data supported the previous MT findings and do not account for the failure to detect RT effects that would support the serial, fixed-order prediction. Similar to experiment 4. the MT data also revealed main effects for action, F(1.11) = 49.54. wz = 0.28, distance, F(l,ll) = 45.81. w* = 0.27, as well as an action X distance interaction, F(l,ll) = 7.55, w2 = 0.01. These values are presented in table 2 and are not of interest to the purposes of this study.
General discussion Information tory processes
processing models that consider, in detail, the preparaof movement typically hypothesize that knowing ‘what
i? D. Lee et al. / Preparing actions and parameters
101
to do’ must be ascertained before ‘how to do it’ can be specified. This hypothesis is derived from theories that consider actions and parameters of action as proceeding in a fixed-order, hierarchical manner. Collectively, and consistently, the present series of experiments offered no support for this prediction. The present results are clearly at odds with a fixed-order model of action and parameter response specification. Rather, the present results are consistent with a variable-order model. With reference to the variables examined here, it appears that when the distance is known in advance, specification of action can occur as fast as when action is known in advance and distance is later specified. That the time to specify one, when the other is known in advance, is the same for actions and parameters of action precludes the possibility that these occur as separate serial stages (cf. Requin et al. 1984; Theios 1973, 1975). Further, that two alternatives specifying neither the action nor the distance still produced latencies equivalent to the other two-choice conditions is additional evidence against the fixed-order hierarchical model. It has not gone unnoticed that the data in these experiments are explained quite well by the Hick-Hyman law (Hick 1952; Hyman 1953). Indeed, when the precue condition RT means from the first four experiments are regressed as a function of the number of SR bits of information (i.e., 0, 1 or 2 bits), the linear equation accounts for 94.4, 91.5, 95.1, and 98.4 per cent of the total variance in each experiment, respectively. Goodman and Kelso (1980), finding no support for Rosenbaum’s (1980) parameter specification model also noted the remarkable similarity of their findings to the Hick-Hyman law. Although given different tasks and procedures, the present findings do not argue against Goodman and Kelso’s conclusion that ‘directly given precues allow the subject to eliminate particular stimulus-response alternatives and prepare those remaining in a more holistic manner’ (1980 : 486-487). The present studies must be considered only as preliminary evidence towards a further analysis of what has largely been an empirically unchallenged notion: that an action must be determined before movement parameters may be specified. Considerably more research using a variety of actions and parameters and tested under different methodologies is required before a more definitive conclusion may emerge.
102
T. D. Lee et (11./ Preparing ac’trons and
parumrtrr.~
References Goodman, D. and J.A.S. Kelso, 1980. Are movements prepared in parts? Not under compatible (naturalized) conditions. Journal of Experimental Psychology: General 109. 475-495. Henry, F.M. and D.E. Rogers, 1960. Increased response latency for complicated movements and a ‘memory drum’ theory of neuromotor reaction. Research Quarterly 31. 448-458. Hick. W.E.. 1952. On the rate of gain of information. Quarterly Journal of Experimental Psychologq 4. 11-26. Hyman. R., 1953. Stimulus information as a determinant of reaction time. Journal of Experimental Psychology 45, 188-196. Kerr. B., 1978. ‘Task factors that influence selection and preparation for voluntary movements’. In: G.E. Stelmach (ed.), Information processing in motor control and learning. New York: Academic Press. pp. 55-69. Klapp. S.T.. 1977. Response programming. as assessed hy reaction time. does not establish commands for particular muscles. Journal of Motor Behavior 9, 301-312. Larish, D.D. and G.A. Frekany, 1985. Planning and preparing expected and unexpected movements: Reexamining the relationships of arm. direction, and extent of movement. Journal of Motor Behavior 17, 168-189. Larish. D.D. and G.E. Stelmach, 1982. Preprogramming, programming and reprogramming of aimed hand movements as a function of age. Journal of Motor Behavior 14. 322-340. Marteniuk. R.G. and C.L. MacKenzie. 1980. ‘Information processing in movement organization and execution’. In: R.S. Nickerson (ed.), Attention and performance VIII. Hillsdale. NJ: Erlhaum. pp. 29-55. Pew. R.W., 1984. ‘A distributed processing view of human motor control’. In: W. Prim and A.F. Sanders (eds.). Cognition and motor processes. Berlin: Springer. pp. 19-27. Requin, J., J.-C. Lecas and M. Bonnet, 1984. ‘Some experimental evidence for a three-step model of motor preparation’. In: S. Kornhlum and J. Requin (eds.) Preparatory states and processes. Hillsdale, NJ: Erlhaum. pp. 259-284. Rosenhaum, D.A., 1980. Human movement initiation: specification of arm, direction. and extent. Journal of Experimental Psychology: General 109. 444-474. Rosenhaum, D.A. and S. Kornhlum, 1982. A priming method for investigating the selection of motor responses. Acta Psychologica 51. 223-243. Sanders. A.F.. 1980. ‘Stage analysis of reaction process’. In: G.E. Stelmach and J. Requin (eds.), Tutorials in motor behavior. Amsterdam: North-Holland. pp. 334-354. Schmidt. R.A.. 1982. Motor control and learning: a behavioral emphasis. Champaign, IL: Human Kinetics Publishers. Theios. J.. 1973. ‘Reaction time measurements in the study of memory processes: theory and data’. In: G.H. Bower (ed.), The psychology of learning and motivation. Vol. 7. New York: Academic Press. pp. 43-85. Theios, J.. 1975. ‘The components of response latency in simple human information processing tasks’. In: P.M.A. Rahhitt and S. Dornic (eds.). Attention and performance V. New York: Academic Press. pp. 418-440.
