Columbia University

Abstract

Language and thought are inextricably entwined. Until the latter half of the 20th century there was little reason to ask whether language and thought could be dissociated, that is, whether thinking could occur without language? That question was raised and answered affirmatively by experiments on animal cognition. The results of such experiments have obvious implications for the evolution of language. To the extent that our non-human ancestors could think without language we can assess the contribution of language to pre-existing cognitive processes.

With that goal in mind, I would like to summarize some experiments I’ve performed on the serial skills of monkeys. Serial skills play a crucial role in intelligent behavior, in particular, during the development of various types of cognitive expertise. In one experiment, monkeys learned, by trial and error, to execute lists composed of 7 arbitrarily selected photographs. The lists were trained with all of the items displayed simultaneously throughout each trial. The location of list items was varied from trial to trial (typically on a touch-sensitive video monitor) to insure that the subjects could not learn a list as a fixed-motor pattern. The task was to respond to each item in a particular order, regardless of its spatial position. To execute a simultaneous chain, subjects had to generate internal cues to define its position in the sequence (that is, think about which item to respond to next) before making each response. Although the probability of executing the entire list correctly by chance was 1/5040, each monkey readily acquired four 7-item lists and showed evidence of increased serial expertise as they learned successive lists.

The following thought experiment illustrates the serial skills needed to execute a 7-item sequence. Imagine trying to enter your personal identification number (PIN) at a cash machine, say 9-2-1-5-8-4-7, on which the positions of the numbers were changed each time you tried to obtain cash. You could not enter that (or any) PIN by executing a sequence of distinctive motor movements, i.e., first pressing the button in the lower right corner of the number pad to enter 9, then the button in the upper middle position to enter 2, and so on. Instead, you would have to search for each number and mentally keep track of your position in the sequence as you pressed different buttons. With the exception that 7 different photographs were used instead of 7 different numerals, this is precisely the problem that the monkeys had to solve when they were trained on 7-item lists. As difficult as that may seem, it would be far more difficult to discover the correct PIN by trial and error. To do so, you would have to recall the consequences of any of the 36 types of logical errors you could make while attempting to produce the required sequence. With the exception that different photographs were used instead of different numerals, this is precisely the problem that monkeys had to solve at the start of training on each of the 4 lists on which they were trained.

Three types of evidence of serial expertise were obtained. (1) There was a progressive decrease in the number of sessions needed to satisfy the accuracy criterion on each list. Subjects needed, on average, 31.5, 17.5, 13 and 12.25 sessions to master Lists 1, 2, 3 and 4, respectively (ranges: 21-55, 11-25, 11-19, and 7-17, respectively). (2) Because food reward was provided only after responding to all 7 items in the correct order, subjects had to determine the correct order in which to respond to the first 6 items of each list solely on the basis of the secondary reinforcement they received following correct responses and errors. (3) Subjects became progressively more proficient at remembering the consequences of correct responses and errors at each position of the sequence. Indeed, solely on the basis of secondary reinforcement they almost succeeded in identifying the first 2 items of their 4th list with the minimum number of logical guesses.

This is the first study to show the development of serial expertise in an animal. Within the framework of human memory systems, the expertise observed in this experiment seems closer to that characterized as declarative knowledge ("knowing that" C comes before E, but after B, etc.) than that characterized as procedural knowledge ("knowing how" to execute a fixed motor sequence). Given that our subjects’ serial expertise cannot be characterized linguistically, it is unclear how a non-human primate represents declarative knowledge. It is clear, however, that procedural knowledge cannot suffice as an explanation of our subjects’ behavior during the execution of a list because they had to generate a different motor sequence on each trial. Another feature of subjects’ serial expertise that falls outside the realm of procedural knowledge is their ability to ascertain an item’s ordinal position with almost maximal logical efficiency.

The sequences that the 4 subjects of this study learned are by far the most difficult lists mastered by non-human primates, including those trained in experiments on the linguistic abilities of ape. It is, however, doubtful that their performance reflects the upper limit of their serial capacity. The ease with which each subject learned 7-item lists and the steady decrease in the number of sessions needed to master new lists suggests that they could learn them in fewer than 13 sessions and that they could also master longer lists. Their performance also shows that the serial expertise for producing arbitrary sequences evolved long before the evolution of language in hominids and that it is possible to investigate mechanisms for planning and executing complex sequences in a non-human primate without the complication of language.

 Conference site: http://www.infres.enst.fr/confs/evolang/