Transposition errors as the tool of motor working memory research Lyakhovetskii V.A., Bobrova E.V., Skopin G.N.* Pavlov Institute of Physiology of Russian Academy of Science * Smolny Institute, Saint-Petersburg State University

Introduction One of the methods of the motor memory research is the analysis of properties of recall of movement sequence to external stimuli. There exists two types of errors: Positional errors Vector errors. So the appearance of transpositions of sequence elements (positions or vectors) is possible. Bock, Eckmiller, 1986

Previously we have shown that distributions of positional and vector errors are distinct for tasks performed by the right and the left hand: the vector coding is predominating in a case of the right hand movements which controlled by the left hemisphere prevalence of positional or vector coding is non- significant for the left hand movements which controlled by both hemispheres. Ляховецкий, Боброва, 2009

Questions Whether the hemispherical specificity for transposition errors exists? Which models can explain transposition errors?

Part I EXPERIMENT

Method The hand of the blindfolded volunteer was moved by experimenter from center through 6 different positions at a sheet of paper A4. The volunteer had to remember and immediately after that to reproduce by pen the sequence of positions. 8 right-handed volunteers. 2 groups: carried out the task by the right hand, and then by the left one (group R), and carried out the task by the left hand, and then by the right one (group L). The task was carried out by each hand within 6 days (6 sheets daily)

Errors of transposition of position Скопин, Боброва, Ляховецкий, Морозова, 2008 Errors of transposition of orientation Agam et al.,

Errors of transposition of position 30 errors 15 errors 20 errors 23 errors 1-6 days7-12 days Begin Left hand, then Right hand Begin Right hand, then Left hand Transpositions between adjacent elements are more probable

Errors of transposition of orientation 43 errors15 errors 44 errors39 errors 1-6 days7-12 days Begin Left hand, then Right hand Begin Right hand, then Left hand

LH begin 23 then RH 15 RH begin 30 then LH 20 RH after LH -> decrease the number of RH errors of transposition of position -> activation of the right hemisphere positional coding system owing to LH sequence acquisition. LH after RH -> decrease the number of LH errors of transposition of orientation -> improvement of the left hemisphere vector coding system owing to RH sequence acquisition. LH begin 44 then RH 39 RH begin 43 then LH 15 Errors of transposition of position Errors of transposition of orientation Conclusions-I

Part II MODELING

Two classes of serial order theories undistributed storage; the content of memory elements is treated as black box explain correct recall after erroneous response, small capacity of wor- king memory, transposition errors RM HR distributed storage (based on artificial neural networks); pay attention to inner representation of memory elements; explain correct recall after erroneous response, small capacity of working memory ? can explain transposition errors 1234 Competitive queuing Chaining RM H R 1234 the next element recall bases on the previous element the weight of each element of memory buffer determines the recall order

Model of sequence recall (according to chaining theories) Associative memory artificial network: stores pairs of sequentially presented objects has fast calculating weighting matrix dont require training performs of self-induced errors Due to ability of this network to make self-induced errors its possible to compare human performance with artificial one. Ляховецкий, Боброва, 2009 ? can explain transposition errors

Model errors of transposition of position Model errors of transposition of orientation Transpositions between adjacent elements are more probable for vector coding. Positional coding by itself cannot explain our psychophysical data – the left hemisphere vector coding system have to be used during the sequence acquisition by both hands.

Conclusions-II Transposition errors can be explained not only by competitive queuing theories, but also by neural network using either vector coding, or both vector and position coding.

Conclusion The data support the conception of the right hemisphere specialization on positional coding and left hemisphere specialization on vector coding The coding system is transferred from one to another hand The results are well described by the model based on the heteroassociative neural network.