Though the recent studies on counting that have investigated either the cost of the coordination or counting span development have led to disconcerting results, we would like to suggest that it would be premature to jettison the resource hypothesis.
If counting is a resource-demanding procedure, a concurrent memory load should have a detrimental effect on counting performance.
Moreover, the effect of memory load in counting should be stronger when counting is more demanding.
As in Experiment 1, Experiment 2 involved adults counting alphabetically to create the most unfavourable conditions for detecting the expected difference in cognitive cost between different types of counting.
The arrays used were similar to those of the Counting Span Task (Case et al.
Errors in saying the number chain while counting and counting times were recorded from an audio-taped recording of each participant's performance.
As far as counting errors were concerned, none of the two load conditions (5-letters and 5-digits) differed from the no-load condition, and the two load conditions did not differ, Fs < 1.
Comparisons were made on counting times averaged across the 10 trials for which the memory load was identical.
To summarize, though no significant increase in the error rates was observed, counting under memory load induced longer counting times independently of the type of material to be maintained (letters or digits).
Our second prediction issuing from the hypothesis that counting is a resource-demanding activity was that the effect of a concurrent memory load should be all the more detrimental when the counting activity is more demanding.
Half of the participants began the task by counting the 18 arrays with the numeric chain and then with the alphabet.
numbers) X 3 (span: 1, 3 and 5 items) ANOVA with repeated measures for all factors was first performed on counting errors.
An ANOVA with the same design was performed on the counting times averaged across the three trials for which the chain used in counting, the type of items to be maintained, and the span were identical.
Expected effects on counting performance related to the type of chain used did not reach significance.
An ANOVA with the same design as that used for counting errors and times was performed on the rate of series correctly recalled after counting.