Exp Brain Res (2003) 149:497–504 DOI 10.1007/s00221-003-1398-4
RESEARCH ARTICLE
Michael O. Russ · Wolfgang Mack · Carina-Raluca Grama · Heinrich Lanfermann · Monika Knopf
Enactment effect in memory: evidence concerning the function of the supramarginal gyrus Received: 22 February 2002 / Accepted: 10 January 2003 / Published online: 4 March 2003 Springer-Verlag 2003
Abstract Experimental behavioral data show that written action descriptions are remembered better when encoded by enacting them compared with merely verbal encoding. To explore this facilitating effect of encoding by performing actions (‘enactment effect’), a functional magnetic resonance imaging (fMRI) study was conducted with n=18 normal subjects. During a learning condition, subjects encoded action phrases like ‘cut the bread’ either by reading aloud or by enacting them. The same phrases plus additional distractors were presented during fMRI scanning, and the task was to decide (yes/no key press) whether a displayed phrase was previously learned or whether it was a new one. Retrieval – independent of encoding type – activated anterior cingulate, SMA, and visual cortex bilaterally. Activations of the inferior frontal and sensorimotor cortex, and the precentral sulcus, were only left sided. The right cerebellum was also activated. The subtraction of the brain activations in the verbal condition from the enactment condition resulted in significant clusters located in middle temporal and inferior parietal left cortical areas, and, on the right side, in superior temporal, postcentral and inferior parietal cortical areas. Most striking were the bilateral inferior parietal activations, covering the supramarginal gyrus (SMG). Therefore it is concluded that SMG may be a
M. O. Russ ()) Klinik fr Neurologie, Klinikum der Universitt, Schleusenweg 2-16, 60528 Frankfurt am Main, Germany e-mail:
[email protected] Tel.: +49-69-63015768 Fax: +49-69-63016842 W. Mack · C.-R. Grama · M. Knopf Institut fr Psychologie, Johann Wolfgang Goethe-Universitt, Frankfurt am Main, Germany H. Lanfermann Institut fr Neuroradiologie, Johann Wolfgang Goethe-Universitt, Frankfurt am Main, Germany
central structure in a neurofunctional explanation of the enactment effect. Keywords Supramarginal gyrus · Memory · Learning by doing · Enactment effect · fMRI
Introduction In psychological research on verbal memory it is a wellestablished finding that encoding simple actions (transitive verb-noun phrases without a subject like ‘cut the bread’) by performing the action improves subsequent memory performance compared to an alternative encoding condition in which action descriptions are merely encoded verbally by reading written items (Knopf 1992b; Engelkamp 1998). Typically, the actions can be performed in a sitting position and the movements required involve predominantly hands and upper limbs. The beneficial memory effect of doing, called the ‘enactment effect’, can be found after performing the action with a real object as well as after pantomiming the action (Nilsson 2000), which is corroborated by different tests of verbal memory including free recall, cued recall, and recognition. Moreover, this effect has been demonstrated in different populations, including the elderly (Knopf 1992a) and patients suffering from memory problems varying in severity (Knopf 1992b) or dementia (Karlsson et al. 1989). What causes the beneficial memory effect of encoding actions by performing them? Obviously, there are many steps of processing between attending to an item and its subsequent recall on request. To find the causes of the phenomenon, all these steps have to be investigated within the whole motor control system (Frith et al. 2000). First, an action phrase has to be read (or heard) and understood, which involves language-mediated semantic analysis. The objects and how they are to be manipulated have to be imagined involving memory for objects and their usage in extrapersonal space. Then, a plan of the suitable sequencing of the subactions has to be set up in
498
order to perform the action on the object as requested. Finally, a motor program steers the execution, the muscles are activated and coordinated and sensory feedback of the movement is registered and compared with the program according to the reafference principle. Subjects’ eyes are fixated to the line when instructed to read only the items, but the eyes are moving freely when real acting or miming of actions is requested. In one research tradition (e.g., Engelkamp 1994, 1998; Nilsson 2000) the most obvious difference between verbal and enactive encoding is studied, the fact that performing an action according to its verbal description requires execution of an ordered sequence of movements, whether with real objects or in mime. If someone is requested to mime ‘cutting a piece of bread’, one has to grasp an imaginary piece of bread with one hand, usually the left one, take a knife in the right hand and move it back and forth as if cutting the bread. It is hypothesized that in addition to verbal (lexical and conceptual) and imaginary information, motor information associated with the production of movement elements making up the required action is critical in producing the enactment effect. The additionally stored motor information facilitates encoding as well as retrieval. During remembering, this motor information is reactivated, and this may produce the superior memory performance for enacted items. For simplicity we call this approach ‘motor information reactivation view’. A second approach to explaining the enactment effect (e.g., Helstrup 1986; Knopf 1992b) focused not on the mere performing of movements as part of the action, but on the representational and motivational components involved in the action realization that are considered to be decisive. We call this approach for simplicity ‘action representation view’. Learning written action phrases by only reading them means that just names of actions are to be remembered. When instructed to mime actions, readiness to act as requested involves a strong selfinvolvement, the formation of an intention to act, an obligatory activation of the action schema, and object knowledge. Action is bound to the existence of a self (Jeannerod 1999). Therefore, ‘enacting’ means cognitive processing at a much higher complexity level than that provided by the primary motor cortex functions. These two views, both based on plausible theoretical reasoning and significant data, can be tested either by behavioral experimental work at a psychological level or by using neuroscience methods, such as electroencephalography or functional imaging, at a brain system level. Brain data do not solve psychological problems per se, but can only be interpreted in correspondence with psychological and behavioral data. But when behavioral data are insufficient to decide a controversy, brain system studies may open up a new window for discussion. Heil et al. (1999) measured event-related brain potentials (ERPs) during recognition of verbally vs. enactively encoded items. In line with the ‘motor information reactivation view’, the question was whether or not reactivation of motor information acquired during encod-
ing of action phrases could be demonstrated by means of EEG measures taken during recognition of the action phrases. A higher negativity during recognition of enactively encoded items at electrodes located over the frontocentral cortex was significant. Additionally, a reliable recognition memory effect existed in the sense that old phrases correctly identified were accompanied by a larger positivity than correctly identified new phrases. This positivity effect was most pronounced within the centroposterior region of the brain and independent from encoding conditions. Due to the restricted spatial resolution of the EEG technique used, interpretation was limited. But the data showed some evidence for the assumption that motor information established during encoding was reactivated in remembering. Nilsson et al. (2000) used PET to test the hypothesis that motor brain areas show increased retrieval-related activity following enacted compared to verbal encoding. They compared verbal, imaginary and enactive encoding by using cued-recall as a dependent variable, with verbs serving as retrieval cues. PET was used to measure brain activity during the retrieval phase. Only activations within motor areas were considered for analysis by constructing a ‘motor mask’ which included cortical motor areas, basal ganglia, and cerebellum. The contrast between the enactment and verbal conditions resulted in a significantly increased activity in the right primary motor cortex. It was concluded that the activity of motor brain areas during remembering of action events may reflect reactivation of motor information acquired during the encoding phase, thus supporting the ‘motor information reactivation view’. But no activation pattern from encoding was obtained by this study. Thus, there was no evidence that activation during retrieval was in fact the ‘reactivation’ of the pattern produced during the encoding phase. Nyberg et al. (2001) addressed directly the question of common activation during encoding and retrieving that the two aforementioned studies left open. Therefore, this study included measurement of brain activity with PET during encoding as well as retrieval of enacted (overt activity) and verbal events. Also included was a control condition involving imaginary motor activity (covert activity) during encoding. Their main concern was to compare the brain activation pattern during retrieval following overt (‘enacted’) versus covered (‘imaginary’) motor activity during encoding, both in direct comparison with their respective encoding patterns. Surprisingly, encoding and retrieval were found to be associated with highly distinct activity patterns, suggesting that different neural systems are operating during memory encoding and retrieval. Nevertheless, a comparatively small overlap on the motor cortex of the left hemisphere between these two patterns was found, which was interpreted by Nyberg et al. in connection with the results of Nilsson et al. (2000) as supporting the ‘motor information reactivation view’ (Engelkamp 1998; Heil et al. 1999). Differential contrasts within the retrieval phase showed significant activations for overtly enacted vs. verbal encoding (inferior frontal, parietal, primary motor and auditory
499
cortex), imaginary enacted vs. verbal (primary motor, premotor, parietal cortex), and overtly vs. imaginary enactive encoding (parietal cortex). All of these activations were found only in the left hemisphere (with one minor exception). Parietal cortex was always pronounced in the activation patterns, and the lateralization to the left side may reflect a bias of the data, because all of the subjects used their right hand only during enactment, and the stimulus presentation as well as the response were made by spoken language. In summary, results of these three studies showed some evidence to support the ‘motor information reactivation view’. But these data should be interpreted cautiously, and not as unequivocally supportive for this view, especially considering the data from Nyberg et al. (2001). Therefore, we tried to replicate some of these findings by using functional magnetic resonance imaging (fMRI), which had not yet been used in brain studies dealing with the enactment effect. We hypothesized that, when subjects are retrieving action items, encoded enactively or verbally, a differential blood oxygenation level dependent (BOLD) effect will be found by contrasting the two conditions. Based on the theoretical position of Knopf (1992b), we tested the following hypothesis: During retrieval of enacted items, the brain region specifically activated in this state is the parietal association cortex of both hemispheres, especially its inferior areas. This a priori hypothesis is based on neuropsychological theory (Heilman and Rothi 1993; Roland 1993; Lezak 1995; Banich 1997; Kolb and Whishaw 1993). The parietal cortex is mediating multimodal processing, somatosensory integration, motor programming, and internal movement representation (Kimura 1977; Mountcastle et al. 1975; Stephan et al. 1995; Sirigu et al. 1996; Gerardin et al. 2000). Functions such as orientation in extrapersonal space, motor attention (Rushworth et al. 1997, 2001), identification of moving body parts, producing and understanding gestures and pantomimed motor acts (Halsband et al. 2001), and the organization of movements (praxis) are mediated here. Tasks such as the naming of actions and spatial relations (Damasio et al. 2001) or retrieval of action words (Tranel et al. 2001) were related to parietal systems as well as knowledge of actions (Martin et al. 1995), imagination with respect to execution, movement ideation (Gerardin et al. 2000), and motor imagery (Decety et al. 1994; Stephan et al. 1995). Lesions to this part of the brain lead to apraxia, agnosia, disorientation in space, left-right confusion, neglect, alexia, and various visuomotor-spatial disturbances. If the facilitating effect of encoding enactively is based on ‘higher order cognitive processing’ in the sense of Knopf (1992b), then the specific activation reflecting the neural basis of the enactment effect should be found in these functionally related parietal areas.
Materials and methods Subjects The 18 participants, 13 females and 5 males, were all white students at the J.W. Goethe University Frankfurt, aged between 20 and 38 years (M=27.4 years, SD=6.0 years), and gave their written informed consent according to the approval of the local ethics committee. They were compensated for their participation. All participants claimed to be right handed (semistructured interview asking the questions of the Edinburgh Handedness Questionnaire), and had no history of drug abuse, or neurological or psychiatric illness. Lefthanders were excluded. Experiment The experiment was run as a within-subjects design with a learning/ encoding phase, and a retrieval phase (during fMRI measurements), both involving two conditions. In the learning condition subjects encoded written German action phrases, which consisted of phrases describing an object and a transitive verb in infinitive form describing the action [e.g., “einen Brief unterschreiben” (to sign a letter), “ein Ei schlen” (to peel an egg), “ein Buch ffnen” (to open a book)]. After encoding, participants performed psychometric tests lasting about 20 min in order to prohibit rehearsal. Twenty-five minutes later, lying in the scanner, subjects were tested for recognition of the phrases learned previously. Differences between the conditions were induced by different encoding instructions. In the enactment condition, subjects read each action phrase presented on a PC monitor for 5 s, and then performed the described action always using both hands to manipulate an imagined object (e.g., taking an imagined boiled egg with the left hand and peeling it with the right hand). In the verbal condition, subjects merely read the action descriptions aloud. All subjects completed both conditions, with the order of condition being counterbalanced over subjects. Each item series comprised 30 randomized items, with item order being randomized for every subject. During recognition, subjects were presented with the target items (encoded by enacting or verbally), mixed with an additional 60 distractor items. The whole series of 120 items was randomized for each subject. Items were presented during the recognition test at a mean rate of 4 s per item, randomly jumping between 2 and 9 s (variable stimulus onset asynchrony). Subjects had to decide on each of the 120 items, if it was old or new, by pressing one of two preassigned buttons with the thumb of the right hand, which was stated as being the dominant hand by all subjects. Image acquisition FMRI was performed using a 1.5-Tesla whole-body scanner (Magnetom Vision) with a standard head coil and gradient booster (Siemens, Erlangen, Germany). Applying an EPI mosaic sequence (Tr/TE=80.7/40 ms, matrix 6464, voxel size 3.443.445 mm, 26 transverse slices, AC-PC orientation), we obtained a series (390 measurements) of blood-oxygenation-sensitive echoplanar image volumes every 3.08 s. Data analysis The functional data were analyzed as ‘event related’ using the SPM99 software from the Wellcome Department of Cognitive Neurology, London, UK, running under Unix and Matlab 5.3 (Mathworks Inc., Sherborn, MA). An event was defined by the beginning of the visual presentation of an action description. All images were realigned (for motion correction), normalized into a standard space (MNI template, Montreal Neurological Institute), and smoothed with a 6-mm full-width-at-half-maximum Gaussian kernel. Low-frequency fluctuations were removed by setting a high-pass filter with 170 s cut-off, and a low-pass filter was set to
500 hemodynamic response function. Post hoc, the 120 events were assigned to their corresponding image acquisition, and these images were then grouped under the appropriate experimental condition. Then, for each subject, a fixed-effect model (within the General Linear Model approach of SPM99) was estimated, and the main condition effects of verbal encoding and enactment as well as the contrasts verbal >enactment and enactment >verbal were calculated based only on events for which the following response was correct. To take the intersubject variability into account, the resulting four tcontrast images (from 18 subjects each) were subjected to a subsequent random effects analysis using a multisubject t-test model. Only activations significant at p
