Elsevier

Acta Psychologica

Volume 130, Issue 1, January 2009, Pages 1-10
Acta Psychologica

Transfer of learning in choice reactions: Contributions of specific and general components of manual responses

https://doi.org/10.1016/j.actpsy.2008.09.008Get rights and content

Abstract

Manifestations of learned skills and knowledge are known to be context-dependent. However, a study of perceptual-motor learning [Tagliabue, M., Zorzi, M., & Umiltà, C. (2002). Cross-modal re-mapping influences the Simon effect. Memory and Cognition, 30, 18–23] reported context-independent transfer of a learned stimulus–response (S–R) mapping to a task in which the mapping is no longer relevant. Although similar results were observed in subsequent studies, these studies also provided an indication that the transfer is context-dependent. The present study investigated the issue of context-dependence of the transfer of a learned S–R mapping. In experiment 1, groups of participants performed choice-reaction tasks with either the same or different response modes (keypresses or joystick movements) in the practice and transfer sessions. Smaller transfer effects were observed for those who switched response mode in the transfer session than for those who did not, indicating that transfer of the learned mapping is context-dependent. However, transfer also occurred for the former group, indicating that the transfer effect is dependent on both general and specific response components. In experiment 2, the same task conditions were examined, but with action effects consistent across the practice and transfer sessions, which were assumed to introduce a contextual feature that was common to the two sessions. The influence of action effects on transfer depended on the practiced response. The results are discussed in terms of feature overlap between the learning and test contexts, and an association network model of learning and response selection.

Introduction

Transfer of newly acquired skills and knowledge to novel task settings is of great interest in educational, occupational, and military research. Training has little utility if trainees fail to apply newly acquired skills outside of the trained context; classroom knowledge should eventually yield practical applications in solving problems in people’s lives, not just problems on a course exam. In spite of this need for generality of skills and knowledge, transfer of learning is known to be context-dependent. In particular, the knowledge and skills are manifested better when the contexts in which learning occurs and is tested are more similar (Bouton, 1993, Godden and Baddeley, 1975, Healy et al., 2005, Healy et al., 2006).

The context-dependence of transfer of learning is explained by way of associations established between some contextual cues and responses during learning (Bouton, 1993). Alternatively, it is explained by the effectiveness of retrieval cues: Contextual factors during learning determine specific encoding operations and thus how learned events are represented in memory; in turn, how events are stored in memory determines how effective retrieval cues present at test are (Tulving & Thomson, 1973). In any case, a change in the context has a potentially harmful consequence for learned skills and knowledge in most cases.

The context-dependence of transfer can be expressed in the following way (Nairne, 2002): Suppose that the practice and test contexts are summarized as vectors of unique features, CP = (X1, …, Xm) and CT = (Y1, …, Yn), respectively. An event E1 of the practice context is associated with the contextual features Xi during the practice. Thus, the effectiveness of the contextual features as retrieval cues for the event E1 can be expressed as f(E1,Xi), where f is some function that represents the associative strength, or weight, of Xi to E1. Thus, f(E1,Xi)=0 if they are not associated, and f(E1,Xi)>0 if they are associated. The activation of E1 in the presence of the contextual feature Yj of the test context is given byU(E1YJ)=i=1mf[E1,S(Xi,Yj)]ki=1mf[Ek,S(Xi,Yj)],where S is a similarity function. Then, the overall activation of E1 is given byU(E1CT)=j=1nU(E1Yj).

Provided that the amount of activation of an event is positively related to the probability of its being retrieved, the likelihood of a contextual feature retrieving a learned event at test depends on two factors: (a) whether the feature matches with some contextual feature present during learning and (b) its associative strength with the event proportional to the sum of its associations with other events present during learning. Consequently, a match of contextual features during learning and testing is the precondition for retrieval of the learned skill and knowledge.

Transfer of learning has been studied in choice-reaction tasks by Proctor and Lu, 1999, Tagliabue et al., 2000. In Proctor and Lu’s study, participants first practiced a task for 900 trials in which they pressed a left key when a stimulus appeared in a right location and a right key when a stimulus appeared in a left location; thus, the task defined a spatially incompatible stimulus–response (S–R) mapping. Subsequently, the participants performed a task in which they responded to a non-spatial stimulus dimension (letter identity), though the stimuli were still presented in left and right locations. Without the prior practice, responses are faster and more accurate when stimulus and response locations correspond than when they do not, a phenomenon called the Simon effect (see Lu and Proctor (1995) and Simon (1990) for reviews). However, after practice with the incompatible S–R mapping, the responses were faster for noncorresponding S–R relations than for the corresponding relations. Tagliabue et al., showed that practice of less than 100 trials with an incompatible spatial mapping was sufficient to eliminate the Simon effect in the transfer session. These observations indicate transfer of the incompatible mapping used in a prior task, despite the fact that the mapping is no longer relevant in a subsequent task.

Tagliabue et al. (2000) explained the transfer of the incompatible mapping in terms of a connectionist model developed by Zorzi and Umiltà (1995). According to Tagliabue et al.’s model, response selection is performed based on response activation via two distinct S–R associations. The first association, a short-term memory (STM) link, connects between the task-relevant feature of a stimulus and a pre-specified response, whereas the second association, a long-term memory (LTM) link, connects between a task-irrelevant, spatial feature of a stimulus and a response spatially corresponding to that feature. Consequently, a spatially corresponding response receives activation from both STM and LTM links, whereas a spatially noncorresponding response receives activation only from the STM link. This is assumed to be the mechanism underlying the Simon effect. However, when participants practice a choice-reaction task with the incompatible mapping prior to the Simon task, they acquire a new STM link that connects between the spatial stimulus feature and a noncorresponding response. Hence, a spatially noncorresponding response now receives activation from two STM links, which eliminates or even reverses the Simon effect, according to the strength of the newly acquired STM link.

In a follow-up study also using less than 100 practice trials, Tagliabue, Zorzi, and Umiltà (2002) examined transfer of the incompatible mapping across different stimulus modalities. When participants practiced the incompatible-mapping task with auditory stimuli and then transferred to the Simon task with visual stimuli, the Simon effect was significantly reduced. Tagliabue et al., concluded, “The effects of practicing a spatially incompatible task cannot be explained by episodic/contextual factors but must be ascribed to a process of spatial remapping that is not modality specific” (p. 22). This conclusion is counterintuitive, provided that transfer of learning is known generally to be context-dependent.

Several studies have followed up Tagliabue et al. (2002) finding (e.g., Proctor et al., 2007, Vu et al., 2003), but the results were not entirely clear with regard to context-dependence of the transfer effect. For instance, Vu et al., showed that when the transfer session used visual stimuli, the transfer effect was significantly smaller for participants who practiced with auditory stimuli than for those who practiced with visual stimuli. However, they also found no significant within-modality transfer effect when the transfer session used auditory stimuli, which made it difficult to unambiguously rule out the possibility that the practice with auditory stimuli was merely insufficient to produce a strong transfer effect (e.g., the learned association was weak; see also Keele et al., 1995, Vu, 2007).

Consequently, the aim of the present study was to demonstrate context-dependence of a learned S–R mapping in choice-reaction tasks. The study closely followed the method used in the studies of Tagliabue et al., 2000, Tagliabue et al., 2002, Vu et al., 2003. However, the manipulated variable was response mode, instead of stimulus modality. Because response mode is part of the task context, the size of transfer effects should also depend on that factor. Furthermore, Vu et al., suggested that the lack of transfer effects for auditory Simon task was due to auditory stimuli yielding a stronger tendency to make a corresponding response than visual stimuli. Hence, the design has an advantage of excluding such a confounding factor by maintaining the effects of visual stimulus modality on the Simon effect.

The first study was divided into two experiments (experiments 1A and 1B) according to the response mode used in the transfer session. The response mode was a standard keyboard in experiment 1A and a joystick in experiment 1B. Because the magnitude of the Simon effect might be affected by the response mode used in the transfer session, where our focus is on the transfer effect, this way of comparing the conditions seems most appropriate. Nevertheless, we provide a supplementary comparison between the two experiments, so that a higher-order interaction between practice and transfer conditions can be examined.

Experiment 2 extended the finding of experiment 1 to another task context in which an additional contextual factor was introduced in the practice and transfer sessions. In particular, regardless of which mode was used, a response produced an action effect that was consistent across the practice and transfer sessions. Previous studies have shown that action effects play an important role in action planning (e.g., Elsner and Hommel, 2001, Hommel, 1993). Thus, that factor can be considered to be a contextual feature. We expected that if consistent action effects are present in the practice and transfer sessions, they increase feature overlap between the contexts and help to maintain the size of the transfer effect when response mode changes.

Section snippets

Experiment 1

Despite the fact that the manifestation of learned skills and knowledge is known to be context-dependent (e.g., Bouton, 1993, Godden and Baddeley, 1975, Healy et al., 2005, Tulving and Thomson, 1973), Tagliabue et al. (2002) argued that transfer of the incompatible mapping to a subsequent choice-reaction task was not. We examined their conclusion by manipulating the response mode used in the practice and transfer sessions.

Experiments 1A and 1B used visual tasks for both practice and transfer

Experiment 2

Experiment 1 showed that transfer of the incompatible mapping was sensitive to a change in response mode and thus context-dependent. At the same time, the experiment indicated that the transfer effect depended on both general and specific response components. Experiment 2 further investigated the contribution of another response variable to the transfer of a newly acquired association.

The correspondence between stimulus and response depends on how the performer interprets, and thus represents,

General Discussion

This study examined context-dependence of transfer of the incompatible mapping. Previous studies (Proctor and Lu, 1999, Tagliabue et al., 2000) showed that the Simon effect is reduced or even reversed after practice with the incompatible-mapping task, indicating transfer of the noncorresponding S–R association. Having shown that the Simon effect with visual stimuli was also reduced after practice with auditory stimuli, Tagliabue et al. (2002) argued that the transfer effect was

Conclusion

The importance of transfer studies of the present type is that, whereas the influence of irrelevant stimulus features on performance is often considered to be robust (hard-wired or overlearned), it is easily overridden by small amounts of practice (Proctor et al., 2007). The present experiments provide clear evidence consistent with the general notion that transfer of learning is context-dependent (e.g., Healy et al., 2005, Tulving and Thomson, 1973). Especially, they suggest that feature

Acknowledgements

This research was supported in part by Grant W911NF-05-1-0153 from the Army Research Office. We thank Ulrich Ansorge, Iring Koch, and Jeroen Smeets for helpful comments on an earlier version of this manuscript.

References (34)

  • R.W. Proctor et al.

    Differentiating types of set-level compatibility

  • J.R. Simon

    The effects of an irrelevant directional cue on human information processing

  • U. Ansorge et al.

    A response-discrimination account of the Simon effect

    Journal of Experimental Psychology: Human Perception and Performance

    (2004)
  • Ansorge, U., & Wühr, P. (in press). Transfer of response codes from choice responses to go-nogo tasks. Quarterly...
  • M.E. Bouton

    Context, time, and memory retrieval in the interference paradigms of Pavlovian learning

    Psychological Bulletin

    (1993)
  • Y.-S. Cho et al.

    Influences of response position and hand posture on the orthogonal Simon effect

    Quarterly Journal of Experimental Psychology

    (2008)
  • B. Elsner et al.

    Effect anticipation and action control

    Journal of Experimental Psychology: Human Perception and Performance

    (2001)
  • B. Elsner et al.

    Contiguity and contingency in action-effect learning

    Psychological Research

    (2004)
  • B.R. Godden et al.

    Context-dependent memory in two natural environments: On land and underwater

    British Journal of Psychology

    (1975)
  • A.F. Healy et al.

    Skill training, retention, and transfer: The effects of a concurrent secondary task

    Memory and Cognition

    (2005)
  • A.F. Healy et al.

    Specificity effects in training and transfer of speeded responses

    Journal of Experimental Psychology: Learning, Memory and Cognition

    (2006)
  • B. Hommel

    Inverting the Simon effect by intention

    Psychological Research

    (1993)
  • B. Hommel

    The cognitive representation of action: Automatic integration of perceived action effects

    Psychological Research

    (1996)
  • B. Hommel et al.

    The theory of event coding (TEC): A framework for perception and action planning

    Behavioral and Brain Science

    (2001)
  • S.W. Keele et al.

    On the modularity of sequence representation

    Journal of Motor Behavior

    (1995)
  • I. Koch et al.

    Verbal response–effect compatibility

    Memory and Cognition

    (2002)
  • W. Kunde

    Response–effect compatibility in manual choice reaction tasks

    Journal of Experimental Psychology: Human Perception and Performance

    (2001)
  • Cited by (22)

    • Dissociating influences of key and hand separation on the Stroop color-identification effect

      2012, Acta Psychologica
      Citation Excerpt :

      Despite the irrelevance of stimulus location, RT is shorter and error rate is higher when the stimulus and response locations correspond than when they do not (e.g., Simon, 1990). For the Simon effect, and studies of stimulus–response compatibility (SRC) proper, for which spatial location is the relevant stimulus dimension, it makes sense to study influences of various response manipulations such as the distance between response keys (e.g., Koch, Schuch, Vu, & Proctor, 2011), whether the alternative keypresses are made by fingers on the same or different hands (e.g., Kornblum, 1965), and whether responses are unimanual movements or discrete keypresses (e.g., Yamaguchi & Proctor, 2009). For the Stroop color-identification task, manual response properties seem of less importance, because the Stroop color stimulus itself conveys no spatial information.

    • Effects of altered auditory feedback across effector systems: Production of melodies by keyboard and singing

      2012, Acta Psychologica
      Citation Excerpt :

      The fact that production relies on sensorimotor coordination is interesting in itself, but perhaps a more compelling question is, how can AAF interference effects inform models of cognitive organization for perception and action? A related issue that has received differing support has to do with the role of different effector systems (i.e., the motor systems responsible for action production): whereas some research suggests that perception–action associations exist at an abstract level of representation that may extend across effector systems (e.g., Cohen, Ivry, & Keele, 1990; Grafton, Hazeltine, & Ivry, 1998; Howell, 2001; MacKay & Bowman, 1969; Palmer & Meyer, 2000), evidence also exists for reduced effects of sensorimotor interactions after switching response mode (Yamaguchi & Proctor, 2009). In addition, some neuroimaging evidence is consistent with the idea that perception–action associations are effector specific (e.g., Buccino et al., 2001).

    • Playing the Simon game: Use of the Simon task for investigating human information processing

      2011, Acta Psychologica
      Citation Excerpt :

      It is also possible to change the mode of responding in the transfer paradigm. Yamaguchi and Proctor (2009) had groups of participants perform the practice and transfer tasks with visual stimuli using either the same response modes (left–right keypresses or joystick movements) in the practice and transfer sessions or different response modes. The transfer effects occurred for those participants who changed response mode in the transfer session as well as for those who did not, but the effects were significantly smaller for those who changed modes.

    • Influence on Simon and SNARC effects of a nonspatial stimulus-response mapping: Between-task logical recoding

      2010, Journal of Experimental Psychology: Human Perception and Performance
    View all citing articles on Scopus
    View full text