The data were collected in 2012.
The ability to classify previously unseen category items into learned categories, sometimes referred to as inductive reasoning or inductive learning, is a fundamental ability with important implications for learning and education. What is the best way to study material in preparation for a classification test? In other words, what is the best study schedule?
Historically, researchers have compared the effects of blocked study — studying members of the same category at once — with effects of intermixed study — interleaving to-be-studied items from different categories. Research has shown that intermixed study is often better than blocked study when it comes to learning new concepts (e.g., [1, 2]; for a recent look, see ). The concept of intermixed study also relates to the well-known phenomenon of the spacing effect, the finding that spacing out study opportunities often leads to greater learning than not spacing them does (see , but also see ). Often, when study trials are intermixed, more time passes between exposures to similar concepts than when trials are blocked (i.e., trials are also spaced).
The data presented here were collected to explore effects of blocked and intermixed study on subsequent classification performance of natural categories — here, different bird species belonging to different bird families. The study was a conceptual successor of another study, which investigated effects of test trials on classification performance . In addition to investigating blocked and intermixed study, we also tested two hybrid study schedules, which combine both blocked and intermixed study elements.
In this study, we also asked subjects to predict their future classification performance using a recent measure of metacognition called the category-learning judgment . This judgment asks subjects to predict future classification performance for new or old items belonging to studied categories. Some researchers believe that these judgments have an important relation to how students choose to schedule their own learning (; for a review of self-regulated learning, see ).
One hundred college students participated at Washington University in St. Louis, MO, a research institution in the Midwestern United States (mean age = 19.2, min age = 18 years, max = 27, SD = 1.6; 34 men, 66 women). Subjects were recruited through an online subject-participation system (Experimetrix) and were awarded cash or course credit for their participation. Subjects were assigned randomly to one of four conditions: blocked, intermixed, blocked-intermixed, and intermixed-blocked. Twenty-five subjects participated in each condition. One subject in the blocked condition did not complete the experiment and was replaced with another.
We used bird illustrations taken from , which originated from the website www.whatbird.com. Each bird (e.g., Baltimore oriole) belonged to a specific family (e.g., oriole). In this study, we used 12 birds, each taken from 12 different families. See [7, 10] for an illustration of the materials. Due to copyright restrictions, these materials cannot be deposited into an online repository; please e-mail the corresponding author to inquire about access to the materials. The experiment was conducted on a computer and programmed with Adobe Flash .
The experiment comprised three phases: (1) study phase, (2) category-learning judgment phase, and (3) classification phase. There were no delays between any phase; the entire experiment was completed in one session. The text of the experiment instructions can be found in the Appendix.
In the study phase, each bird was presented for 8 s, and the name of the family to which the bird belonged appeared underneath. In the blocked condition, subjects studied six birds from each family; each group of six was studied separately for each of the 12 families. The six birds of a family were selected randomly from the 12 birds of that family from the stimulus set; families were presented in a random sequence. In the intermixed condition, subjects were presented with six birds from each of the 12 families in random sequence. In the blocked-intermixed condition, subjects studied three birds from each family, blocked by category, and then the remaining three birds from each family were presented in a random sequence. Last, in the intermixed-blocked condition, subjects studied three birds from each of the 12 families in a random sequence, then saw the remaining three birds from each family presented in a blocked sequence. Each subject saw 72 birds total. In all cases, the order of species was random, and the assignment of birds to study (and subsequent test) was also random.
During the study phase, after a bird appeared on the screen, subjects were instructed to press the space bar on the computer keyboard when they felt that the bird had been learned fully. The bird remained on the screen for the remainder of the 8 s after the space bar was pressed. If the subject did not press the space bar within 8 s, the bird disappeared from the screen, and subjects were asked to click a button indicating why they did not press the space bar. The options were (1) learned, meaning the bird was learned, but the space bar was not pressed; (2) not learned, meaning that the bird was not learned; and (3) not paying attention, meaning that the subject “zoned out” and forgot to press the space bar. After the space bar was pressed or a button was clicked, a 500-ms blank screen followed before the next bird was presented. There were no additional pauses or stops during the study phase.
In the category-learning judgment phase, subjects were asked to judge how well they believed they would be able to classify new birds belonging to studied families on an upcoming classification test. All family names appeared on the screen beside sliding scales that ranged from 8% (indicating chance, i.e., ≈ 1 ÷ 12) to 100%. Subjects predicted their future classification performance by using these sliding scales. Judgments were self-paced.
In the classification phase, subjects were presented with the remaining six birds from each of the 12 categories in random order. Beside each bird was a list of all the studied bird families. Subjects attempted to classify the bird appearing on the screen by selecting the correct family name. Classification attempts were self-paced and no feedback was provided. After the subject attempted to classify all 72 birds, the experiment was completed and the subject was debriefed.
The experiment was conducted in a quiet research lab. Subjects participated individually. Five experimenters (including Cecilia Votta) collected the data. Subjects were told that they could ask the experimenter if they had any questions.
The research was conducted under the oversight of the Washington University in St. Louis Institutional Review Board. The data are anonymous and contain no demographic or other identifying information.
(3) Dataset description
The data file is named “Data.xlsx.”
The data are processed. They have been formatted from what was captured by the experiment initially.
Format names and versions
The data are saved as an open XML spreadsheet file (Excel file; .xlsx).
K. Andrew DeSoto (graduate student at the time of data collection) and Christopher Wahlheim (postdoctoral fellow) supervised data collection. Cecilia Votta (undergraduate student) and four research assistants in the Aging, Memory, & Cognitive Control Laboratory at Washington University in St. Louis, MO, collected the data.
The data file is annotated in English.
The data have been deposited under a CC-BY open license, “reuse with attribution.”
The data were published on figshare on December 12, 2015.
(4) Reuse potential
These data are of potential interest to cognitive psychologists, educational psychologists, computer scientists, or other researchers interested in modeling human learning. They describe the order in which subjects learned specific stimuli belonging to different categories and provide information on both common ways of scheduling learning (i.e., blocked and intermixed schedules) as well as on an underexplored type of study schedule (hybrid schedules). The data also reflect performance on a classification test of natural categories, a topic of both theoretical and applied interest.
In addition to having reuse potential for learning researchers, the data may also be useful for metacognition researchers. Subjects’ category-learning judgments and subsequent classification performance allow the assessment of measures of metacognitive monitoring such as calibration and resolution .
Last, the data presented here are suitable for inclusion in meta-analyses investigating any of the topics described above. We encourage researchers to use this dataset for teaching or collaboration purposes, too.
The authors declare that they have no competing interests.