Various sources of distraction during analogical reasoning

Abstract

Reasoning by analogy requires mapping relational correspondence between two situations to transfer information from the more familiar (source) to the less familiar situation (target). However, the presence of distractors may lead to invalid conclusions based on semantic or perceptual similarities instead of on relational correspondence. To understand the role of distraction in analogy making, we examined semantically rich four-term analogies (A:B::C:?) and scene analogies, as well as semantically lean geometric analogies and the matrix task tapping general reasoning. We examined (a) what types of lures were most distracting, (b) how the two semantically rich analogy tasks were related, and (c) how much variance in the scores could be attributed to general reasoning ability. We observed that (a) in four-term analogies the distractors semantically related to C impacted performance most strongly, as compared to the perceptual, categorical, and relational distractors, but the two latter distractor types also mattered; (b) distraction sources in four-term and scene analogies were virtually unrelated; and (c) general reasoning explained the largest part of variance in resistance to distraction. The results suggest that various sources of distraction operate at different stages of analogical reasoning and differently affect specific analogy paradigms.

Keywords: Analogy; Distraction; Mapping; Reasoning.

Fig. 1

Sample item of the Four-term Analogy. The A:B::C:D? problem (top) to solve: puddle:rain boot::hot pot:? Response options (bottom) are the following: ​an oven mitt is the correct response, a gas burner is a semantic distractor related to C, ​a frying pan is a categorical distractor related to C, a basket is a perceptual distractor related to C, an umbrella is a semantic distractor related to B, a boot is a categorical distractor related to B, a chess pieces are a perceptual distractor related to B, and a rubber glove is a relational distractor. The order of response options was fully random across all trials in the study

Fig. 2

Example items of Scene Analogies. a A cross-mapping item. On the left, an elderly man is explaining something to a group of students, so he plays the role of a leader. On the right, an elderly woman is assisted by a young man who plays the role of a leader (so, mapping the two elderly persons together leads to an error). The four response options at the bottom of the panel included the correct response (the young man), a distractor (the elderly woman belonging to the same category of people as the man pointed to in the first scene), and two irrelevant (i.e., non-distractor) objects (a woman passing by with a child and a bench). b A non-cross-mapping item. Response options comprise the correct response (an outdoor tap) and three irrelevant (non-distractor) objects. c A cross-mapping item. On the left, a woman is distracting a man biker who, as a result, spills gasoline on a bag. On the right, a woman is being distracted by a man neighbor while watering her plants, as a result spilling the liquid on the laundry. The four response options included the correct response (the man), a distractor (the woman) and two irrelevant (i.e., non-distractor) objects (a watering can and the laundry). d The c pair is reversed and cross-mapping is absent. Response options comprise the correct response and three irrelevant objects. For all items, the order of response options in the response set was random for each participant

Fig. 3

a An exemplary item of Geometric Analogies and b an RAPM-like item (not an actual item); the correct response (dashed rim) is missing and has to be selected out of four/eight options in the case of Geometric Analogies/RAPM (not depicted in the example)

Fig. 4

Percentage of error choices in Four-term Analogies, computed as the proportion of the number of times each error option was committed over all errors committed. The bars indicate 95% confidence intervals

Fig. 5

a A Venn diagram showing the amount of variance in the total error rate on Four-term Analogies (FTA) predicted uniquely and jointly by the distractor error rate in Scene Analogies (CM), Geometric Analogies (AR), and Raven’s Advanced Progressive Matrices (Gf). b An analogous diagram showing the amount of variance in CM predicted uniquely and jointly by FTA, AR, and Gf. Null values are not shown. The numbers representing the amounts of variance were calculated using the variance partitioning method based on the R² values of the regression models presented in Table 2. See text for description