In Experiment 1, meaningless target-detection stimuli were used to remove the specificity of detection stimuli and explore whether target detection can continue to affect recognition in such circumstances. Furthermore, an FA condition was added to explore the effect of target detection on memory retrieval and to calculate the difference between the FA condition and the target and distractor trials related to recognition performance. Contrasting FA and DA can clarify the influencing mechanism of target detection on recognition.
Methods
Participants. Overall, 33 undergraduate students (14 males; mean age = 19.85 years; standard deviation [SD] = 2.22) from Fujian Normal University participated in this experiment, for which they received a reward of ¥15. The sample size was decided a priori to match Huang and Meng (2020). In their experiments, the average effect size for the rate of correct identification of old words was ηp2 = 0.15, which transforms into an effect size of f = 0.42. In Experiment 1, an α of .05 and a sample size of 33 were determined to equate to a power of .95 to detect an effect of f ≥ 0.26 (all power estimates were sourced from G*Power 3.1.9.2; Faul et al., 2007). Data for two participants were excluded because of a low target-detection rate, and a further two were excluded as a result of a low recognition rate. All procedures were approved by the Fujian Normal University Institutional Review Board, and all participants provided verbal informed consent.
Materials
Memory Materials. One hundred ninety-two double-character words were selected from a Chinese lexical database (Sun et al., 2018). The frequency of the selected words ranged from 0.01 to 8.02 per million to ensure the same frequency of words, and the average frequency of the words was 1.13 ± 1.57 per million. The words were generally neutral in meaning. Thirty-two individuals, who did not participate in the experiment, rated the words’ respective pleasantness and level of arousal using five-point scales (1 = extremely unpleasant/very calm; 5 = extremely pleasant/very excited). This assessment showed that the mean scores for pleasantness and arousal were 3.16 ± 0.42 and 1.92 ± 0.48, respectively. The full set of words was randomly divided into two groups: half to the FA and half to the DA condition. In both conditions, 64 words were presented in both the learning and test stages (representing ‘old’ words), and 32 were presented in the test stage only (representing ‘new’ words). In the test stage, for both types of words, half were presented with the target, and half were presented with the distraction stimulus. This design yielded four trial types (an old word with target stimulus, old word with distractor stimulus, new word with target stimulus, and new word with distractor stimulus). There were no significant differences among different conditions regarding the words’ level of pleasantness, level of arousal, frequency, and number of strokes (pleasantness: F(1,192) = 0.019, p = 0.891; arousal: F(1,192) = 0.402, p = 0.527; word frequency: F(1,192) = 0.428, p = 0.514; number of strokes: F(1, 192) = 0.111, p = 0.739). The presentation of the stimuli was balanced across the participants.
An additional set of 24 double-character words were used as practice items.
Detection Stimulus. Red and green squares with sides of 1 cm in length.
Procedure
The experiment was conducted in a standard psychological-behaviour laboratory. E-prime 2.0 was used as the experiment programme. The computer used in the experiment was a Lenovo ThinkPad with a 19-inch Liquid Crystal Display and a resolution of 1,280 × 1,024 pixels. The words were presented on the computer screen in white over a black background. The experiment comprised an FA and DA condition, each of which comprised a study phase, distractor task, and memory test. The following is an example of the DA condition:
(1) Study phase: 64 words were randomly presented in the centre of the screen, one at a time. Each word was presented for 1 s. Participants were asked to remember the word by visualising the described object.
(2) Distractor task: Participants performed an arithmetic-based distractor task for two minutes.
(3) Test phase: A set of 96 words was randomly displayed, one at a time, in the centre of the screen, and a coloured square was presented concurrently with each word. Each coloured square was located 1 cm above each word. Each trial began with both the word and square visible on-screen; after 100 ms, the coloured square disappeared, and after 700 ms, the word disappeared; this left a blank screen that remained for 2,200 ms (Fig. 1). Participants were asked to report whether each word was old or new. If the word had previously appeared in the study stage (i.e., was an ‘old’ word), participants pressed the ‘9’ button with the index finger of their right hand; if the word was new, participants pressed the ‘0’ button with the middle finger of their right hand. Concurrently, participants were asked to attend to the colour of the square presented above each word, pressing the ‘1’ key with the left middle finger if the square was the target square and not pressing any key (other than ‘9’ or ‘0’, depending on whether the word was old or new) if it was the distractor square. The target’s colours (red or green) and the distractor squares were balanced between the participants. The participants were told that both tasks were equally important and were asked to respond to both tasks as quickly and accurately as possible after each set of stimuli appeared. As a result of the short presentation duration, participants were allowed to respond after the words had disappeared.
The presentation of the FA stimuli was identical to the DA stimuli, except that the participants were asked to respond only to the words and ignore the square above them. In addition, the participants performed the FA and DA conditions in an order balanced across the sample.
Results and Discussion
Referring to the study of Huang and Meng (2020), the present study analysed the recognition accuracy of new and old words and the reaction times for such recognition. Moreover, the discrimination index (d′) and the criterion of judgment (C) were analysed. All statistical analyses were conducted using IBM SPSS Statistics for Windows, version
20.0 (IBM Corp., Armonk, N.Y., USA).
Results for target detection. Overall, in the test phase, the correct identification rate of the target coloured square was 98.12%. Meanwhile, the correct identification rate of target squares accompanying old words was 97.95%, and the average response time was 903.50 ± 292.90 ms. In comparison, the correct identification rate of target squares accompanying new words was 98.28%, and the average response time was 1030.60 ± 291.62 ms. A paired sample t-test showed that trials in which the target square accompanied an old word were associated with significantly higher reaction times than trials in which the target square accompanied new words [t(28) = −5.82, p < 0.001, Cohen’s d = −1.08, 95% CI [−1.54, −0.62]]. There was no significant difference between the two in regard to accuracy [t(28) = 0.47, p = 0.639].
Recognition results. Table 1 presents the recognition results.
Table 1 Recognition Results for Experiment 1
|
|
Correct identification of old words
|
Correct rejection of new words
|
Discrimination index (d’)
|
Response bias (C)
|
|
Target
|
Distractor
|
Target
|
Distractor
|
Target
|
Distractor
|
Target
|
Distractor
|
FA
|
ACC
|
0.71
(0.17)
|
0.73
(0.14)
|
0.84
(0.13)
|
0.84
(0.12)
|
1.95
(0.98)
|
2.04
(1.07)
|
0.34
(0.59)
|
0.34
(0.55)
|
RT
|
678
(137)
|
700
(143)
|
757
(154)
|
764
(146)
|
DA
|
ACC
|
0.77
(0.14)
|
0.65
(0.17)
|
0.77
(0.14)
|
0.84
(0.12)
|
1.75
(0.72)
|
1.73
(0.75)
|
0.02
(0.49)
|
0.43
(0.59)
|
RT
|
897
(235)
|
978
(235)
|
1087
(242)
|
1001
(227)
|
Note. Standard deviations are presented in parentheses, d′= Z (correct identification rate) − Z (false response rate); C = −1/2[Z (correct identification rate) + Z (false response rate)] (Huang & Meng, 2020).
ACC: accuracy; DA: divided attention; FA: full attention; RT: reaction time.
The correct identification rate of old words and rate of correct rejection of new words (see Fig. 2) were respectively submitted to a 2 × 2 analysis of variance (ANOVA) that featured attention condition (FA vs. DA) and trial type (target vs. distractor) as factors. The results follow.
Correct identification of old words. Regarding recognition rate, the main effect of attention condition was not significant [F(1,28) = 0.14, p = 0.712], but the main effect of trial type was [F(1,28) = 4.75, p = 0.038, ηp2 = 0.15]. Importantly, the interaction between them was significant [F(1,28) = 18.70, p < 0.001, ηp2 = 0.40]. Further simple-effects analysis showed that, in the DA condition, the rate of correct identification of old words was significantly higher in target trials than in distractor trials [F(1,28) = 25.56, p < 0.001, ηp2 = 0.48]; there was no significant difference between such trials in the FA condition [F(1,28) = 0.46, p = 0.502]. Meanwhile, the correct identification rate of old words in target trials was slightly higher in the DA condition than in the FA condition [F(1,28) = 4.09, p = 0.053, ηp2 = 0.13], whereas the rate of correct identification of old words in distractor trials was significantly lower in the DA condition than in the FA condition [F(1,28) = 6.66, p = 0.015, ηp2 = 0.19] (see Fig. 2).
Regarding reaction time, the main effect of the attention condition was significant [F(1,28) = 42.74, p < 0.001, ηp2 = 0.60]; significantly higher average reaction times were observed in the FA condition than in the DA condition. Further, the main effect of the trial type was also significant [F(1,28) = 20.40, p < 0.001, ηp2 = 0.42]. Importantly, the interaction between attention condition and trial type was significant [F(1,28) = 7.11, p = 0.013, ηp2 = 0.20]. Further simple-effects analysis showed that, in the DA condition, the average reaction time for old words in target trials was significantly faster than that for old words in distractor trials [F(1,28) = 17.32, p < 0.001, ηp2 = 0.38]. Meanwhile, no significant differences in this regard were observed in the FA condition [F(1,28) = 3.60, p = 0.068].
Correct rejection of new words. Regarding the rate of correct rejection of new words, the main effect of attention condition was not significant [F(1,28) = 3.91, p = 0.058, ηp2 = 0.12], and neither was the main effect of trial type [F(1,28) = 2.84, p = 0.103, ηp2 = 0.09]. However, the interaction between them was significant [F(1,28) = 4.23, p = 0.049, ηp2 = 0.13]. Further simple-effects analysis showed that, in the DA condition, the rate of correct rejection of new words was significantly lower in target trials than in distractor trials [F(1,28) = 6.59, p = 0.016, ηp2 = 0.19], but not in FA [F(1,28) = 0.01, p = 0.936]. In addition, the DA condition showed a significantly lower correct rejection rate for target trials than the FA condition [F(1,28) = 9.11, p = 0.005, ηp2 = 0.25], but there was no significant difference between DA and FA for distractor trials [F(1,28) = 0.01, p = 0.935] (see Fig. 2).
Regarding reaction time, the main effect of attention condition was significant [F(1,28) = 60.73, p < 0.001, ηp2 = 0.68], as was the main effect of trial type [F(1,28) = 12.02, p = 0.002, ηp2 = 0.30]. Moreover, the interaction between them was significant [F(1,28) = 13.02, p = 0.001, ηp2 = 0.32]. Further simple-effects analysis showed that, in the DA condition, the average response time for new words in target trials was significantly slower than that for new words in distractor trials [F(1,28) = 19.30, p < 0.001, ηp2 = 0.41]; however, there was no such difference in the FA condition [F(1,28) = 0.21, p = 0.647].
Results of signal detection theory. d′ and C were subjected to the same repeated-measures ANOVA. The results showed that, for d′, the main effects and interaction of attention condition and trial type were not significant (ps > 0.05).
Meanwhile, for C, the main effect of attention condition was not significant [F(1,28) = 2.55, p = 0.122], but the main effect of trial type was [F(1,28) = 5.06, p = 0.032, ηp2 = 0.15], as was the interaction between them [F(1,28) = 7.62, p = 0.010, ηp2 = 0.21]. Further simple-effects analysis showed that, in the DA condition, target trials were associated with a significantly lower response bias than distractor trials [F(1,28) = 14.17, p = 0.001, ηp2 = 0.34], whereas there were no significant differences between these variables in the FA condition [F(1,28) = 0.004, p = 0.949]. Finally, comparing the DA and FA conditions, for target trials the response bias for old words was significantly lower in the DA condition than in the FA condition [F(1,28) = 10.92, p = 0.003, ηp2 = 0.28], whereas there was no significant difference in this regard for distractor trials [F(1,28) = 0.75, p = 0.393].
Discussion
Experiment 1 replicated the results of Huang and Meng (2020). The average correct identification rate of old words was significantly higher in target trials than in distractor trials. Signal-detection analysis showed no difference in discrimination index between the target and distractor trials, whereas the response bias was lower in the former than in the latter. In Experiment 1, the target symbol was changed from the ‘+’ used in Huang and Meng’s (2020) study to a meaningless coloured square. However, the effect of target detection on memory retrieval remained, indicating that the symbol used is not the reason target detection affects memory retrieval.
In addition, an FA condition was added to Experiment 1 to explore the differences between the DA and FA conditions. It was consequently found that the main effect of the attention condition was not significant. Moreover, for the discrimination index, the main and interaction effects of attention condition and trial type were not significant (ps > 0.05). This result is consistent with existing findings that memory retrieval is, in some sense, mandatory and protected (Anderson et al., 1998; Craik et al., 1996; Craik et al., 2018; Greene et al., 2021; Hou et al., 2022). However, when we compared target and distractor trials regarding the average rate of correct identification of old words, significant differences were observed between them. Moreover, we compared performance in the target and distractor trials with those in the FA condition, and the results were entirely unexpected. For target trials featuring old words, no significant differences were found between the DA and FA conditions; however, the average rate of correct identification of old words in distractor trials was significantly lower in the DA than in the FA condition. Meanwhile, the average rate of correct rejection of new words in target trials was significantly lower in the DA condition than in the FA condition. However, there were no significant differences for distractor trials. It seems that, by directly comparing the results of the DA condition with those of the FA condition, the real effect of attentional resources on memory retrieval could not be observed. Dynamic attention resources influence memory retrieval. This effect is reflected in changes in response bias. More critically, the effect of target detection on memory retrieval is based on the influence of the target and on the effect of the distraction. These results will be further explained in the general discussion.
In Experiment 1, the ratio of old words to new words was 2:1. However, a previous study has found that the more old words included, the easier participants find it to make correct ‘old’ responses (Koop & Criss, 2016). Thus, the change in the response bias in Experiment 1 may have originated from the proportion of old and new words rather than target detection. Therefore, in Experiment 2, we adjusted the ratio of new to old words to further explore the effect of target detection on recognition.