The simultaneous recognition of multiple words: A process analysis

Abstract

In everyday life, recognition decisions often have to be made for multiple objects simultaneously. In contrast, research on recognition memory has predominantly relied on single-item recognition paradigms. We present a first systematic investigation into the cognitive processes that differ between single-word and paired-word tests of recognition memory. In a single-word test, participants categorize previously presented words and new words as having been studied before (old) or not (new). In a paired-word test, however, the test words are randomly paired, and participants provide joint old-new categorizations of both words for each pair. Across two experiments (N = 170), we found better memory performance for words tested singly rather than in pairs and, more importantly, dependencies between the two single-word decisions implied by the paired-word test. We extended two popular model classes of single-item recognition to paired-word recognition, a discrete-state model and a continuous model. Both models attribute performance differences between single-word and paired-word recognition to differences in memory-evidence strength. Discrete-state models account for the dependencies in paired-word decisions in terms of dependencies in guessing. In contrast, continuous models map the dependencies on mnemonic (Experiment 1 & 2) as well as on decisional processes (Experiment 2). However, in both experiments, model comparison favored the discrete-state model, indicating that memory decisions for word pairs seem to be mediated by discrete states. Our work suggests that individuals tackle multiple-item recognition fundamentally differently from single-item recognition, and it provides both a behavioral and model-based paradigm for studying multiple-item recognition.

Keywords: Cognitive modeling; Continuous models; Discrete-state models; Recognition memory.

Fig. 1

Illustration of the general recognition theory for paired-word recognition using the mean group-level posterior parameters of Experiment 2. Ellipses represent the different stimulus types: black for two new words, gray for one old word paired with one new word, and light gray for two old words. Dashed lines indicate the response criteria that separate the respective responses

Fig. 2

Extended two-high threshold model for paired-word recognition. Abbreviations for response categories and stimuli: NN—both words new; NO—left word new and right word old; ON—left word old and right word new; OO—both words old. The trees for pair types ON and OO are not shown. For a full description of the parameters, see Supplementary Material

Fig. 3

Median percentages and interquartile ranges of given responses (‘Observed’) to the respective pair types and the predicted frequencies of the two-high threshold model (2HTM) and general recognition theory (GRT) for single words (panel A) and paired words (panel B). N—word not studied (new); O—word studied (old); NN—both words new; NO—left word new, right word old; ON—left word old, right word new; OO—both words old. Outliers are not depicted

Fig. 4

Hit and false-alarm rates for single-word trials (single) and paired-word trials split into words appearing on the left (paired left) and the right (paired right) side of the screen for each condition (blocked/mixed). Error bars reflect the respective standard errors. *** = p < .001

Fig. 5

Median percentages and interquartile ranges of given responses (‘Observed’) to the respective four pair types and the predicted frequencies of the two-high threshold model (2HTM) and general recognition theory (GRT). NN—both words not studied (new); NO—left word new, right word studied (old); ON—left word old, right word new; OO—both words old. Outliers are not visualized