Rat sequence learning and observation.

In Chapter 2 it was argued that rats are capable of learning about stimuli via observation (Jacoby & Dawson 1969, McQuoid & Galef 1992) and that they can learn sequences of stimuli and responses asocially (Hulse & Campbell 1975, Reid 1994, ), but the question remained as to whether rats could learn a sequence of stimuli or responses following observation of a trained conspecific. Whiten (1998) provided the only evidence of observational sequence learning to date which suggested that animals (chimpanzees) were capable of learning about a two item

sequence following observation of a human demonstrator and that this learning may occur primarily through associative processes (D'Amato & Colombo 1991). Therefore it was expected that the same processes would be seen to be in operation in rat observational sequence learning.

Experiments 6.1 & 6.2 sought evidence of observational sequence learning using a standard two action test incorporating both location and directional information. By placing observers and demonstrators in the same operant chamber during observation sessions , a free-operant procedure was carried over to testing. Based upon evidence from Terrace (1991), who suggested that in order for rats to learn a sequence the individual sequence elements must be available within a rats current repertoire, observer rats in Experiment 6.1 were pretrained on each lever and in each direction in isolation prior to observing a competently trained demonstrator executing a sequence. Results showed no evidence of observational sequence learning by comparing sequence discrimination ratios. However, when data was analysed in terms of the individual responses that made up the sequence, analysis revealed a recency effect (Shimp 1976). That is, if observers were exposed to the sequence RDLU they were reliably more likely to press the lever left up than observers who had been exposed to a RULD sequence. The effect of recency was, however, only apparent during one testing session therefore no firm conclusions based upon this result can be made and interpretation should be approached with extreme caution. Scent cues deposited on the terminal sequence element may have been responsible for the recency effect, but it was not clear why a comparable effect was not found for right lever responding.

Although Terrace's (1991) suggestion that pretraining animals prior to observation would promote evidence of sequence learning as individual sequence elements would be now present in their repertoire, it was thought that the biases in responding that emerged during training may have resulted in a carry over effect during testing, in part leading to a failure to find evidence for the reproduction of observed sequences. Therefore Experiment 6.2 dispensed with pretraining the

observer rats. Results of Experiment 6.2 demonstrated that, regardless of the exclusion of pretraining, observers who saw a sequence demonstrator were no more likely to reproduce sequences of responses than a control group that had been exposed to demonstrators that merely collected food pellets from the food tray. In fact the level of reproduction of observed sequences was minimal (2% of total responding). Analysis of the individual responses that made up the sequences, revealed a recency effect and this time the effect was consistent across sessions.

It was thought that experiments 6.1 & 6.2 may have failed to provide evidence of observational sequence learning because the sequences RULD and RDLU were not sufficiently distinct. Thus in experiment 6.3, observers were exposed to demonstrators pressing LR or RL before testing. The primary function of the fixed trial procedure employed in experiment 6.3 was to enhance the salience of the sequence and to decrease the delay between observation of a sequence and testing. This procedure also helped to obliterate any effects that scent cues might have on observer behaviour following the recency effects found in Experiments 6.1 and 6.2. This new procedure meant that observers could manipulate their own levers immediately following a single demonstration of the sequence by the demonstrators. However, even following these changes to the procedure observers reproduction of demonstrated sequences was found to be so low that it did not warrant statistical analysis. Analysis of perseverating sequences on the other hand e.g. LL or RR sequences, yielded an effect of primacy which was in direct contrast to the preceding two experiments. If an observer had been exposed to a LR sequence they were found to make a greater proportion of LL responses compared with a group that was exposed to a RL sequence. The emergence of a primacy effect was argued to be due to observer rats paying more attention to the demonstrators first response and so neglecting to pay attention to any second response the demonstrator made, in the absence of scent cues. Shimp (1976) argued that making changes to the retention interval between demonstrator responding and observer responding may have a profound effect on the emergence of both recency and primacy effects. Having a small retention interval increased the

possibility of the emergence of a primacy effect as in Experiment 6.3 whilst a longer retention interval was found to promote a recency effect as in Experiments 6.1 and 6.2. When making adjustments to the fixed trial procedure, for use in future studies, it would be appropriate to examine the effects of making changes to the time interval between demonstrator sequence execution and observer lever manipulation in an attempt to eradicate the effects of primacy and recency thus promoting sequence learning.

Overall experiments 6.1, 6.2 and 6.3 provided no evidence of observational sequence learning and there are a number of suggestions as to why this might be the case. It could be argued that the failure to find any effect of observation might indicate that rats are incapable of observational sequence learning, however, on the basis of only three experiments it is not possible to state this categorically. Although some adjustments are required to this procedure to ensure that observers pay full attention to the latter parts of the demonstrators response, possibly by requiring observers to orient towards a particular location on the wall separating the observation and demonstration compartment prior to observer levers being inserted into the observation chamber, or making changes to interval between demonstrator responding and observer responding, this procedure should be viewed as a significant innovation in experimental research on social learning in animals. In addition, rats are known to possess poor eyesight, particularly in terms of their visual acuity ( 1 cycle per degree. Rats, 60 cycles per degree. Humans) and so may not be an ideal species with which to investigate learning of this kind. Maybe studies of this nature should be confined to using birds or apes who are known to have eyesight similar if not better than our own. However, on a more positive note, the evidence for learning about responding to single sequence items is more convincing and is supported by other experimental data (Heyes, Ray, Mitchell & Nokes, under review). The new fixed trial procedure, however, requires some refinement which can only take place following continued use. However this procedure is a significant innovation within the field of observational learning.