Memory trace strength and response biasing in short-term motor memory*

Transcription

1 Memory & Cognition 1975, Vol. 3 (1), Memory trace strength and response biasing in short-term motor memory* GEORGE E. STELMACH and J. A. SCOTT KELSO University ofwisconsin, Madison, Wisconsin Two experiments, which attempted to create differential memory trace strengths in a response biasing paradigm, were performed. After the presentation of the criterion location, an interpolated target was presented which was either ± 40 deg from the criterion. The 8's task was to attend to both targets and recall each when instructed. The first experiment involved strengthening the criterion trace via repetition (0, 5, or 14 rep.) while the second involved providing additional feedback via visual, auditory, and heightened kinesthetic cues. In the initial experiment, a Repetition by Response Biasing interaction revealed that repetition systematically reduced error shifts at recall. The second experiment found that, m the combined feedback and visual conditions, response biasing was reduced. it seems feasible to suggest that both studies successfully manipulated memory trace strength which appears to be one determiner of error shifts at recall. I' Directional error shifts in the' recall' of a positioning response following the presentation of an interpolated movement have been common in recent short-term motor memory (STMM) literature. Evidence has been presented which indicates that such error shifts are in the direction of the interpolated movement (Patrick, 1971; Pepper & Herman, 1970; Craft & Hinrichs, 1971; Laabs, 1973; Stelmach & Walsh, 1972, 1973). Thus, if the latter is of greater intensity or extent than the criterion, recall error is influenced in a positive direction. Similarly,if the interpolated movement is of lesser intensity or extent, constant error at recall is shifted in a negative manner. At present, two views have been expressed to account for this phenomenon, both ofwhich rely to some extent on the concept of assimilation (Helson, 1964). Pepper and Herman (1970) have postulated a theory of trace interaction in which the memory traces of the criterion and interpolated movements interact to yield a memory trace that is a combination of both. The recall response of the criterion is, therefore, made with reference to the altered trace representation and the directional error at recall is seen as an assimilation effect. Laabs (1973), on the other hand, views the reproduction of the criterion as being made in reference to an "average" or "central" movement, which he terms adaptation level (Helson, 1964), and to the criterion memory trace. Changes in adaptation level as a result of interpolated motor activity are thus seen as responsible for shifts in recall errors. As yet, empirical data which differentiates between these two positions has been tthis research was supported by Research Grant MH from the National Institute of Mental Health and by Grant NE-G-oO from the National Institute of Education awarded to the first author. The research was conducted in the University of Wisconsin-Madison Biotron, a controlled environment research facility, supported by the National Science Founcation and the University of Wisconsin. Requests for reprints should be sent to Dr. George E. Stelmach, Motor Behavior Laboratory, University of Wisconsin, 2000 Observatory Drive, Madison, Wisconsin difficult to obtain. Recent research has focused rather on characteristics of the interpolated response which influence recall error shifts in an effort to shed some light on the foregoing theories. For example, Pepper and Herman (1970), Herman and Bailey (1970) and Craft and Hinrichs (1971) have all demonstrated reproduction errors proportional to the magnitude of the interpolated Stelmach and Walsh (1972, 1973) have examined, in turn, the duration of time spent at an interpolated location and the temporal placement of the interpolated movement within the retention interval, and found both variables to have potent effects on shifts in recall error. In the first of these studies it was assumed that, since S remained at the interpolated location during the retention interval, the interpolated memory trace remained stable, while the memory trace associated with the criterion movement decayed over time. Similarly in the second study, it was argued that the later the interpolated movement was presented within a retention interval (i.e., the closer to recall), the stronger its trace would be relative to the criterion trace when S made his recall responses. Response shifts at recall in both studies were therefore interpreted in terms of.the relative strengths of the two memory traces. The present study reports two experiments which further examined the relative trace strength interpretation. The first experiment attempted to strengthen the memory trace of the criterion movement by means of repetition. The Ss made either 0, 5, or 14 repetitions to a criterion location prior to the presentation of an interpolated location (Adams & Dijkstra, 1966). The second experiment manipulated trace strength by increasing the information feedback to the S via additional auditory, visual, and heightened kinesthetic cues. This latter technique has been used.effectively by Adams, Goetz, and Marshall (1972) and Stelmach (1973) to reduce forgetting in STMM. A 58

2 reduction in error shifts as a function of the strength of the criterion memory trace was predicted in the present experiments on the basis of the relative trace strength hypothesis. EXPERIMENT I Method Subjects. Thirty Ss (15 males and 15 females) between the ages of 18 and 26 years were obtained through the Financial Aids Office at the University of Wisconsin. Each was randomly assigned to one of three experimental conditions with the restriction that an equal number of both sexes appeared in each condition. All Ss were paid $1.50 for participation. Apparatus. The apparatus consisted of a manual lever identical to one described previously by Stelmach and Walsh (1972). The S was instructed to move the free-moving, near frictionless lever at a steady rate to variable target locations which were defined by stops inserted by E. Positioning and reproduction responses were made in a right to left manner and errors were recorded to the nearest 0.5 deg (1 deg =4 mm of displacement). Design. A 3 by 2 by 2 by 5 design using three independent groups (repetition conditions) with repea ted measures on the last three factors (retention interval, response biasing, and targets) was utilized to examine accuracy of kinesthetic recall. Five criterion movements were combined with a positive and negative biasing target which in turn was paired with a 5-sec and 20-sec retention interval, giving a total of 20 trials which were randomized for all Ss. Procedure. The S assumed a seated position facing the lever apparatus. Standardized instructions were read to each S who was then given one familiarization trial. From his position behind the lever apparatus, E administered testing procedures and recorded S's recall estimates. Each S was informed that it was necessary to remember two movements in each trial. For the zero repetition condition (0 REP), S was simply presented the criterion target (CT = 50, 70, 110, 115, and 130 deg) followed by the biasing target (BT) which was systematically varied from the criterion by plus 40 deg (positive) or minus 40 deg (negative). Either immediately or after a 20-sec retention interval which commenced when S returned from the CT, S recalled both targets (CT and BT) in order of presentation. The 5 repetitions (5 REP) and 14 repetitions (14 REP) conditions were identical to the aforementioned except that S, after receiving the criterion presentation, repeated that MEMORY TRACE STRENGTH IN RESPONSE BIASING 59 movement either 5 or 14 times. On each occasion S remained at the target location for 2 sec and then returned to the 0 deg starting position for an identical period of time. Following the designated number of repetitions, S was presented the BT (which was called MOVE TWO by E) and returned to the starting position for the duration of the retention interval, after which he recalled the CT and BT in order of presenta tion. During all trials S maintained a firm grip on the lever handle, but was allowed to release and rest during the 20-sec intertrial interval. The E administered all verbal commands on the basis of light signals which were provided by a programmed Lafayette eight-channel timer. Results Mean constant and absolute errors for criterion recall with accompanying standard errors for the experimental variables of repetitions, retention intervals and response biasing are presented in Table I. Analysis of constant error revealed that only the main effects of response biasing and target locations were significant F(1,27) =68.48, P <.01 and F(4,108) =31.48, P <.01, respectively. In addition, only the Repetitions by Response Biasing interaction was found significant F(2,27) =8.39, p <.01. Inspection of this interaction indicates that response biasing (positive and negative) reduces as a function of repetition. To further examine the magnitude of this reduction, difference scores were calculated by subtracting the negative biasing error from the positive biasing error for each of the three repetition conditions. The means in degrees for the 0, 5, and 14 REP conditions were 7.40, 3.40, and 2.37, respectively, and clearly illustrate a reduction in error shifts as a function of repetition. While absolute error is not as interesting as constant error in response biasing studies, due to the fact that these scores do not reflect directional shifts in error, they are of value in assessing the behavioral effects of repetition, retention interval and targets, and are therefore included in this analysis (see Table 1). The main effects of repetitions and retention intervals were significant F(2,27) = 12.39, p <.01 and F(1,27) = 6.50, Table 1 Means (in Degrees) of Repetition Conditions, Retention Intervals, and Response Biasing for the Recall of the Criterion Targets Retention Interval Response Biasing Target Locations Experimental Group Conditions Means osec 20 Sec Positive Negative ORep 5 Rep 14 Rep CE* Sx** AEt Sx CE Sx AE Sx CE Sx AE Sx "Constant error **Standard error of the mean fabso[ute error

3 60 STELMACH AND KELSO p <.05, respectively. Further analysis of the repetitions effect utilizing Newrnan-Keuls' method revealed that the orep condition was significantly different from the 5 REP and 14 REP condition (p <.01) which were not statistically significant from each other. Neither response biasing, targets or any interaction effects reached a significant level. Inspection of constant and absolute errors for the recall of the interpolated movements revealed that both sources of error yielded no significant main or interaction effects. EXPERIMENT II Method Subjects. Fifty-five Ss between the ages of 18 and 30 years were obtained from the Financial Aids Office at the University of Wisconsin. Each S was randomly assigned, with the restriction that there were six female and five male Ss in each of the five experimental groups (N =11). All Ss were paid $1.50 for participation. Apparatus. The lever apparatus in this experiment was identical to the one utilized in Experiment I, except that some modification was necessary to allow manipulation of visual, auditory and kinesthetic feed back information (Stelmach, 1973). In order to manipulate visual cues, a sliding shutter (35 x 30 em) was used which when opened enabled S to view his positioning movements. Attached to the axis of the lever was a 360-tooth 62 pitch gear, directly in front of which was a 5-cm platform. A removable plastic card was mounted on the platform with one end resting on the gear. With the card in place, a rapid clicking sound occurred as the lever was moved, thus providing S with auditory cues. Also mounted on the lever axis alongside the gear was a 10-cm diam steel drum. Over the drum was a 23 x 2.5 cm leather belt attached at one end to a coiled metal spring which in turn was attached to the lever frame. The other end of the belt was fastened to a metal crank, which, when rotated, provided a tension on the lever handle, creating a torque on the lever of 854 g. As in Experiment I, all movements originated at a O-deg location and all timed events were controlled by a programmable timer. Design. A 5 by 2 by 5 independent groups design (feedback conditions) was utilized with repeated measures on the last two factors (response biasing and target locations). The five criterion targets used in Experiment I were paired with a positive and negative biasing target (±40 deg with respect to the criterion), thereby making 10 pairings which were randomly ordered for each S. For the entire experiment, S received 40 trials, each of the 10 pairings being replicated four times. For the pupose of analysis, replications were collapsed for constant and absolute error inspection while variable error was calculated as the standard deviation of each S's constant error over replications. Procedure. Following a brief familiarization period with the lever apparatus, standardized instructions were read to each S who then received one practice trial. For all conditions a trial consisted of two movement presentations, a retention interval (20 sec), and two recall estimations. For each criterion presentation, S displaced the lever to his left from a O-deg starting position until it made contact with a stop peg which defined a movement location. The S remained there for 2 sec, and on the command "release," re-moved his hand to a resting position while E returned the lever to the starting position. After a period of 10 sec, S was presented the biasing target in the same manner. Then, following the retention interval, S recalled both movements in order of presentation. In the control condition, Ss made blind positioning movements to criterion and biasing locations following the established paradigm. Since Ss received no visual (+V), auditory (+A), or heightened kinesthetic (+K) feedback information, this condition was labeled ~ VAK. Feedback information in the remaining four exper rnental conditions was manipulated only on the criterion presentation and recall, while the interpolated movement and its recall were performed as in the control condition (-VAK). Thus, in the visual condition (+V), S could see his hand and arm as he displaced the lever during criterion presentation and recall. Similarly in the auditory condition (+A), S was able to hear his lever movements. Heightened kinesthetic cues were provided in the +K condition; while in the +VAK conditions S8 received feedback information via the manipulation of all three modalities. Results The mean constant error for the five feedback conditions, two response biasing conditions and five target locations are reported in Table 2. The analyses of signed errors revealed that feedback was not a significant variable F(4,50) = 1.44, P >.05. However both the main effects of reponse biasing and targets were significant F(I,50) = 94.96, P <.01 and F(4,200) = 27.68, p <.01, respectively. The comparison of most interest, the Feedback by Response Biasing interaction was significant, F(4,50) =5.19, p <.05. In order to obtain a clearer view of this interaction, difference scores were calculated between the positive and negative biasing errors for each feedback condition. The means in degrees for the +VAK, +V, +A, +K, -VAK conditions were 1.57, lao, 4.23, 5.27, and 4.01, respectively, and clearly demonstrate reduced error shifts in the +VAK and +V conditions compared to the remaining feedback conditions. Variable error calculated as the standard deviation of each S's constant error across the four replications is also represented in Table 2. It can be seen that responses in the feedback condition which had all three modalities available (+VAK) were the most consistent. Analysis of these mean errors revealed that the main effect of feedback was significant F(4,50) = 12.94, P<.01. Using Newman-Keuls' method, it was found that +VAK was significantly different from +K, +A and - VAK (p <.01), while +V differed significantly from +K (p <.05), +A and-yak (p <.01). No other main or interaction effects reach significance. Examination of the absolute error in Table 2 revealed that the +VAK and +V feedback conditions have smaller mean errors than the three other conditions. Analysis of these scores revealed a main effect of feedback F(4,50) =15.49, P <.01. Post hoc analysis of means using Newman-Keuls' method, indicated that the +VAK and +V conditions differed significantly from +A, +K, and -VAK conditions, (p <.01), but not from each other. The only other main or interaction effect that reached significance was target locations F(4,200) = 4.69, p <.01. The constant errors for the recall of the interpolated recall targets were -0.67, -0.50, -1.81, -2.10, and

4 MEMORY TRACE STRENGTH IN RESPONSE BIASING deg for the +VAK, +V, +K, +A, and -VAK conditions, respectively, and were not significantly different from each other. The means for positive q.86 deg) and negative (-0.36) biasing targets did differ significantly FO,SO) = P <.01. A significant main effect of.target locations F(4,200)=30.18,p <.01, was also obtained. In addition, the Groups by Biasing interaction reached significance F(4,50 = 4.32, P <.01. Inspection of this interaction revealed that the positive targets were quite consistent across feedback conditions, while the negative targets varied considerably from condition to condition in a rather inconsistent pattern. DISCUSSION The present study was the first to attempt to directly manipulate memory trace strength within a response biasing paradigm. Previous investigations have manipulated the temporal characteristics of the interpolated movement but have made no direct efforts at strengthening either trace. These experimental manipulations produced greater error shifts at recall leading to a relative decay state interpretation. In terms of the present study, this hypothesis would predict that response biasing would be increased or decreased depending on the relative strengths of the criterion and interpolated traces. The stronger, or more stable the criterion trace compared to the interpolated trace, the less error shift at recall, since a stronger trace is assumed to be more resistant to the influence of an interpolated The results from the two experiments reported here are in agreement with this view and strongly suggest that memory trace strength is a factor in the degree of response biasing. In the first experiment, it was found that increased repetition of the criterion response led to a reduction in response biasing and absolute error at recall. Similarly, in the second experiment, increased feedback on the criterion movement in the form of visual, auditory, and heightened kinesthetic cues produced less error shifts at recall. In addition absolute and variable error were decreased as a function of augmented feedback. Adams' (1971) closed-loop theory asserts that the strength of a given movement trace is dependent on the amount of practice and feedback impinging upon it. Recently, Adams, Goetz, and Marshall (1972) and Stelmach (1973) have provided evidence to support this notion; both investigations finding that augmented feedback (+VAK) provided markedly reduced recall errors in comparison to a condition were S had minimal proprioceptive information. However, neither Adams' theory or the aforementioned studies, delineated the relative contributions of each modality to trace strength although it was assumed that all input was of equal importance. With regard to Experiment II, however, the data suggest that vision may be more important in strengthening the criterion trace, since kinesthetic and auditory cues by themselves did not seem to affect response biasing to any great degree. This finding agrees with Adams and Goetz (1973) who found that visual cues were dominant in regulating the perceptual trace, Groups Biasing Positive Negative Targets 50 Deg Table 2 Means (in Degrees) of Feedback Conditions and Response Biasing for the Recall of the Criterion Targets 70 Deg 110 Deg 115 Deg 130 Deg Feedback Conditions +VAK +V +A +K -VAK CE AE VE CE AE VE CE AE VE CE AE VE CE AE VE Mean SE Mean , ,43 SE Mean , , SE Mean , SE , Mean SE Mean 2, , SE , , Mean 0, , SE Mean -0, , ,45 SE Note-CE = constant error, AE = absolute error, VE = variable error, and SE = standard error of the mean.

5 62 STELMACH AND KELSO and raises the question of whether a change is needed in Adams' (1971) original construct that all feedback channels are equally involved in the control of While the two existing theories in STMM (Laabs, 1973; Pepper & Herman, 1970) have addressed themselves only to blind positioning acts, it is of interest to view them in light of the feedback data in Experiment II. As it now stands, Pepper and Herman's theory is unable to account for the dominant effect of vision on recall, since it postulates a memory trace which is incremented or decremented solely by kinesthetic and proprioceptive information. Such interpolated stimuli appear somewhat ineffective in altering a criterion trace which has been laid down with vision as a contributing modality. Laabs' theory, on the other hand, is less specific than Pepper and Herman's, since it postulates changes in adaptation level (which presumably can occur across modalities) to account for.assimilation effects in STMM. Under this notion, any interpolated motor activity which changes the adaptation level (AL) can be expected to produce a shift in recall error. While it is unclear as to how AL is derived, the present data would indicate that interpolated activity is less able to affect AL when the memory traces involved are laid down by the visual mode. This interpretation appears to invite some speculation as to the locus of assimilation effects in STMM. By establishing the memory trace using the visual modality, the effect of an interpolated act in the kinesthetic mode appears to have little influence. On the other hand, in spite of attempts to strengthen the memory trace via the kinesthetic modality an interpolated act of the same stimulus mode induces a large biasing effect. Although more research is required to identify the mechanism involved, the present data suggest a peripheral rather than a central locus for assimilation. Support for this interpretation is provided in recent studies by Craft and Hinrichs, (1971) and Craft (1973) in which Ss were informed prior to (pre cueing) or after (post cueing) the presentation of movements as to which was to be recalled. Presumably if response biasing was a peripheral phenomenon, the pre cueing instruction would not reduce recall error shifts. This notion was supported, since constant error was the same in both conditions. Aside from response biasing effects, Laabs (1973) has addressed the question of the appropriate index of decay. As the memory trace decays S is presumed to become less consistent in his reproduction responses, leading Laabs to advocate variable error (VE) as the proper index of decay. Since there were no replications of targets in Experiment I, it was not possible to explore the effects of repetition on VE. On the other hand, Experiment II did allow an examination of VE and indicated that it was reduced as a function of augmented feedback. This finding appears to support the notion that VE is a sensitive indicator of trace strength and is in contrast to the recent work of Marteniuk (1973) where VE was not considered capable of indexing trace decay or discriminating between movement cues in STMM. In conclusion, recent research has established that response biasing can be influenced by the magnitude of the interpolated act (Craft & Hinrichs, 1971), the duration of time spent at an interpolated location (Stelmach & Walsh, 1972) and the temporal occurrences of interpolated activity within the retention interval (Stelmach & Walsh, 1973). In addition, the present study has indicated that memory trace strength is a potent variable. Unlike the two existing theories of STMM, all of these findings can be accommodated easily by a relative trace strength interpretation. This hypothesis appears to have some potential in explaining assimilation effects in STMM. REFERENCES Adams, J. A. A closed-loop theory of motor learning. Journal of Motor Behavior, 1971, Adams, J. A. & Diikstra, S. Short-term memory for motor responses. Journal of Experimental Psychology Adams, J. A., & Goetz, E. T. Feedback and practice as variables in error detection and correction. Journal of Motor Behavior Adams. J. A., Goetz, E. T., & Marshall, P. H. Response feedback and motor learning. Journal of Experimental Psychology (3), Craft, J. L. A two process theory for the short-term retention of motor responses. Journal of Experimental Psychology, 1973, Craft, J. L., & Hinrichs. J. V. Short term retention of simple motor responses: Similarity of prior and succeeding responses. Journal of Experimental Psychology Helson, H. Adaptation level theory, New York: Harper & Row, Herman. L. M., & Bailey, D. R. Comparative effects of retroactive and proactive interference in motor short-term memory. Journal of Experimental Psychology, 1970, Laabs, G. E. Cue effects in motor short-term memory. Journal of Experimental Psychology, 1973, 100, Marteniuk, R. G. Retention characteristics of motor short-term memory cues. Journal of Motor Behavior, 1973, Patrick. J. P. The effect of interpolated motor activities in short-term motor memory. Journal of Motor Behavior, 1971, 3, Pepper. R. 0. & Herman. L. M. Decay and interference effects in short-term retention of a discrete motor act. Journal of Experimental Psychology, Monograph, 1970, 83. Number 2. Stelmach. G. E. Feedback: A determiner of forgetting. Acta Psychologica, 1973, 37, Stelmach, G. E. & Walsh, M. F. Response biasing as a function of duration and extent of positioning acts. Journal of Experimental Psychology, 1972, 92, Stelmach, G. E. & Walsh. M. F. The temporal placement of interpolated movements in short-term memory. Journal of Motor Behavior (Recieved for publication April 9, 1974, revision received June 10, 1974.)