(Editor:   Maureen McCarthy)


Trial and error learning in humans can be studied by giving students a task that is analogous to that faced by white rats in mazes (e.g., Olton & Samuelson, 1976). Historically, students have participated in this learning task by tracing finger mazes while blindfolded (e.g., Corey, 1931; Barker, 1932). The study located at this site is a computer-based analog to these traditional maze learning tasks.

The maze used in this study consists of a start box, 16 chambers, and a finish box. An overhead view of the maze is given in Figure 1.

Sample overhead image from Maze study
Figure 1

The task is to navigate through the maze. Participants are provided with a straight-ahead view from their current position in the maze. The view shows the open doors or paths that can be chosen. Participants use the mouse to click on the directional arrows in order to choose their path. For example, Figure 2 shows the view from the start box. Note that there are just two directional options active (straight ahead and right) from the start box. On subsequent choices, only viable options are active. When participants reach a dead end, the only active choice will be a backward movement.

Sample image of start position of Maze study
Figure 2


There are two dependent measures for this study. Timing is the first measure and reflects the elapsed time for completion of the maze. If participants return to the start box before entering the finish box, the timing is not restarted.

The second dependent variable is error. The numbers of errors are recorded during each trial. Each time an incorrect movement, which results in a deviation from the correct path is made, it counts as one error. Each subsequent movement along that incorrect path is not counted as an additional error.

The experiment requires that participants complete the maze 15 times. Data can be plotted for each of the 15 trials and should reflect a decrease in the time to complete the maze.

Data can also be aggregated. Typical learning studies average a series of trials into one block. In experimental design, a "blocking variable" is used to capture unwanted variance from a nuisance variable. The term "block" in this study refers simply to a set of trials. Choice of block size is simply a matter of desired resolution. In the study of human and animal learning, block sizes of 1 (i.e., trial by trial report) are apt to produce jagged learning curves. Block sizes that are too large may miss an important temporal aspect to learning. So the trick is to find a proper block size. We have used 3 and that seems to work quite well. These blocks can then be plotted and the resulting graph produces a smoother representation of learning. In this study, it would be appropriate to average trials in groups of 3 and this would result in 5 data points.

Data Format and Download

Data are downloadable in three formats (XML, Excel spreadsheet format, and comma delimited for statistical software packages like SPSS). Definitional information for each of the labeled columns are shown in Figure 3.

Partial sample data image from Maze study
Figure 3

The first column provides the participant ID number. The third column provides the class ID for which a participant's data is affiliated. Gender (column 2) and age (column 4) are self-explanatory. The date of participation is located in column five. Columns labeled T1T-T15T reflect time for completion of trials 1-15. Columns labeled T1E-T15E reflects errors for completion of the maze for trials 1-15.

Data Analyses

In this repeated trials study, it would be appropriate to conduct a repeated measure ANOVA for time and/or errors in the maze task. The ANOVA can be conducted using all 15 trials or the 5 suggested blocks.

Applications and Extensions

This study can be used to illustrate classic learning experiments in which rats or people explore novel situations and make decisions based on stimuli that produce desired outcomes. Early learning studies were conducted at the basic level of analysis and provided a foundation for cognitive processing (see Hilgard, 1948). Newer models suggest that learning is multifaceted and may be influenced by multiple factors (Ahsen, 2000). Nevertheless, analyses of the results of this study provide students with understanding of the basic elements or building blocks of cognition.


Ahsen, A. (2000). Image and maze: Learning through imagery functions. Journal of Mental 
	Imagery, 24(1 &2), 1-60.

Barker, R. G. (1932). Factors influencing transfer between finger mazes. Journal of General 
	Psychology, 6, 115-132.

Corey, S. M. (1931). An inexpensive finger maze. American Journal of Psychology, 43, 502.

Hilgard, E. R. (1948). Theories of learning. New York: Appleton-Century-Croft.

Olton, D. S., & Samuelson, R. J. (1976). Remembrance of places passed: Spatial memory in rats. 
	Journal of Experimental Psychology: Animal Behavior Processes, 2, 97-115.