In Experiment 1, participants were exposed to familiar everyday objects (e.g., tables, chairs) before engaging in a recognition practice phase whereby they performed old–new recognition on a subset of the studied objects. The key question was whether recognition practice of an object impairs memory for unpractised objects that share only the same shape (Rp–Shape), or only the same colour (Rp–Colour) with the practised (Rp+) objects. If both shape and colour are represented in object memory, then both properties should drive competition effects in memory, and thus RIF would be present both for unpractised objects sharing only shape (Rp–Shape) and for those sharing only colour (Rp–Colour) with the practised objects. The relative magnitude of RIF in the Rp–Shape and Rp–Colour conditions would further indicate the strength or contribution of these two types of property at retrieval.
Sharing both shape and colour features with the practised object (e.g., Rp–Both) was also expected to produce significant RIF, possibly of similar magnitude as Rp–Colour and Rp–Shape objects (see Reppa et al., 2013). Finally, RIF was expected for Rp–Neither objects as they shared category (e.g., chairs) but neither colour nor shape with the practised objects, which has been previously shown to lead to significant RIF for visual objects (e.g., Maxcey & Woodman, 2014; Reppa et al., 2017). However, it seemed reasonable to expect that RIF magnitude might be lower for Rp–Neither objects compared to objects sharing both category and visual features (i.e., Rp–Colour, Rp–Shape, and Rp–Both).
Forty Swansea university students over the age of 18 years took part in the study in exchange for course credit. One group of 20 participants were allocated to the recognition-practice group (seven males and 13 females), and a different group of 20 participants (four males, 16 females) were allocated to the control group (no recognition practice). All reported normal or corrected-to-normal vision and normal colour vision. All were native English speakers and naïve to the purpose of the experiment.
Apparatus and stimuli
Experiment 1 was run on a Dell OptiPlex GX520 computer connected to a 15.5-in. LCD monitor. Stimulus presentation, trial randomisation, and recording of responses and response times were controlled via E-prime (version 2.0).
Object images were taken from Art Explosion 750,000 and the World Wide Web. The images were modified using Strata 3D pro and Adobe Photoshop and fit within a square (not visible during presentation) of approximately 10 × 10 cm with a resolution of 71 dpi. When viewed from a distance of 60 cm, the objects did not exceed 9.52 × 9.52° of visual angle.
Figure 2 shows the 88 object images used in Experiment 1 (40 targets, eight distractors used in the practice phase only, and 40 distractors used in the test phase only). The stimuli were pictures of everyday objects belonging to one of four different categories: Tables, Chairs, Lamps and Vases – although these names were never mentioned to the participants.
In each object category (e.g., tables) there were two object sets (set 1 and set 2; Fig. 2, Panel A, five leftmost columns). For each object set there was a single Rp+ object for all participants in the recognition practice group. Corresponding Rp–Shape objects shared the exact same shape with Rp+ objects but differed in colours. Rp–Colour objects differed in shape from the Rp+ objects but had the same colours and texture as them. Rp–Both objects shared exactly the same shape and the same colours as the Rp+ objects, but the colours of the parts were re-assigned. For instance, if the Rp+ object was composed of a dark blue table top and light brown legs, then the Rp–Both object would be identical in shape to the Rp+ object, but the table top would be light brown and the legs would be dark blue. Finally, the Rp–Neither objects shared neither shape nor colour with Rp+ objects.
Once all the target (old) objects were created, distractor (new) objects (Fig. 2, Panel A, five rightmost columns) were made for each target object. Distractor objects were identical to targets in terms of colour (and colour combinations), as well as being similar to the targets in terms of overall shape configuration, and different from the targets in the shape of their individual parts. For the Rp+ objects, two distractor objects were created using the aforementioned constraints; one distractor object that was used as a distractor during the test phase only (Fig. 2, Panel A), and another that was used during the recognition practice phases (Fig. 2, Panel B).
For the practice group, a repeated-measures design was used manipulating Practice with two levels: practised versus unpractised categories, and Item Type with five levels: Rp+ (practised objects), Rp–Shape (objects sharing shape with Rp+ objects), Rp–Colour (objects sharing colour with Rp+ objects), Rp–Both (objects that shared both the same shape and the same colour with the Rp+ objects), Rp–Neither (objects that did not share shape or colour with Rp+ objects).
For the control group a repeated-measures design was used manipulating only item type with five levels: Rp+, Rp–Colour, Rp–Shape, Rp–Both and Rp–Neither. As there was no recognition practice for the control group, those terms were irrelevant. However, control participant performance served as a baseline for the assessment of RIF. Furthermore, control group performance on each item type would help ensure that any effects of Item Type (which were of key interest here) were not caused by effects of baseline discriminability of each item type in memory. The dependent variable for both groups was target discriminability, expressed in terms of A’ (Snodgrass, Levy-Berger, & Haydon, 1985).
Participants were seated individually in a quiet room approximately 60 cm from the computer monitor. Participants in the recognition practice group completed a single study phase, three recognition practice phases separated by filler tasks, and a test phase. Control participants completed the study phase, a filler task and the test phase, but not the recognition practice phase. The filler task took as long to complete as the recognition practice phases did for the practice group participants.
There was a single study phase, the same for both groups, where the 40 target objects appeared one at a time at screen centre, in a random order. Participants studied each object for 5 s and were told to expect a later memory test of the objects shown during the study phase.
Recognition practice phases
Only the practice group participants completed the recognition practice phase. Consistent with previous recognition practice experiments, participants completed three practice phases (e.g., Maxcey & Woodman, 2014; Reppa et al., 2017). In each of the practice phases, four of the studied objects were practised (two objects from two different object categories, e.g. two tables and two chairs) together with four distractor objects – one distractor object for each Rp+ object. Therefore, in each practice phase eight objects appeared (one at a time): four Rp+ objects and four distractor objects, yielding a total of 24 practice trials per participant across the three practice phases. In each practice phase, objects appeared at screen centre individually, in random order, with phases separated by filler tasks. During recognition practice, participants indicated whether they believed each object had been studied or not in the earlier phase (the study phase), by pressing either the Q or P key on a QWERTY keyboard. Half of the participants responded ‘Yes’ by pressing P with their right hand and ‘No’ by pressing Q with their left hand, with the response key reversed for the other half of the participants. There were equal numbers of old and new stimuli, and thus equal numbers of correct ‘Yes’ and ‘No’ answers. The practised categories for half of the participants (e.g., tables and chairs) were the unpractised categories for the other half of participants (e.g., lamps and vases), and vice versa. Correct responses were followed by a ‘Correct’ message on the screen for 1 s, and incorrect responses were followed by an ‘Incorrect’ message on the screen for 1 s concurrently with a 500-Hz beep sound. Incorrect trials were not replaced. There was no time limit for responding.
For the Practice group participants, filler tasks were used in-between the three practice phases and in-between the last practice phase and the test phase. The first and second filler tasks (between the first and the second practice phase, and between the second and third practice phase, respectively) lasted for 2 min and the third (between the third practice phase and the test phase) lasted for 5 min. Filler tasks required participants to list as many words as they could for each letter of the alphabet, for a range of categories (e.g., girls’ names, animals, capital cities) with a different category used after each practice phase. The Control group participants completed the same practice tasks, but without the recognition practice element.
The two participant groups completed the same test phase. All objects that were presented in the study phase, as well as all their associated distractors (see Fig. 2, Panel A) were presented individually and in a random order at screen centre. The task was identical to that of the recognition practice phase with participants indicating whether they had seen each object during the study phase or not using the same response keys as recognition practice group participants did during the practice phase. Fast and accurate responding was emphasised, and corrective feedback was provided in the same way as during the practice phases.
Recognition practice success
Recognition practice success was measured in terms of high discriminability of the practised (Rp+) studied objects against the distractors used during practice. Recognition practice in Experiment 1 was successful, with target objects being successfully discriminated from distractor objects (Hits: M=.92, SD=.09; False alarms: M=.11, SD=.15, A’: M=.94, SD=.08).
Test phase analyses
Mean accuracy measures in Experiment 1 are shown in Table 1. The mean A’ per Rp condition in Experiment 1 is shown in Fig. 3. The comparisons between Nrp and each of the Rp conditions were planned comparisons, based on the predictions outlined in the introduction. The Bonferroni correction was applied to all planned comparisons, and Cohen’s d effect sizes are reported for each.
Control group analysis
A one-way repeated-measures ANOVA examining the effect of Item Type (Rp+, Rp–Colour, Rp–Shape, Rp–Both, and Rp–Neither) on A’ scores showed a significant main effect, F(4, 19)=11.97, p=.003, ηp2 =.39. Simple effects analysis to examine the main effect of Item type showed that Rp–Both yielded higher A’ scores than Rp+, (p=.04) Rp–Colour (p=.02) and Rp–Shape items (p<.001). There were no other significant differences.
Practice group analysis
To estimate within-participants RIF and facilitation, a 2 (Practice: practised vs. unpractised categories) × 5 (Item Type: Rp+, Rp–Colour, Rp–Shape, Rp–Both, and Rp–Neither) repeated-measures ANOVA was first carried out on A’ scores from the practice group. There was a significant main effect of Practice, F(1, 19)=4.46, p=.05, ηp2 =.19, with higher A’ scores for the unpractised compared to the practised object categories. There was also a significant main effect of Item Type, F(4, 76)=5.43, p<.001, ηp2=.22. The interaction was only marginally significant, F(4, 76)=2.20, p=.07, ηp2 =.10. Planned comparisons against the Nrp baseline showed significant RIF for Rp–Shape, t(19)=2.74, p=.02, d=.56 and Rp–Colour objects, t(19)=2.24, p=.03, d=69. There was no significant RIF for Rp–Both objects, t(19)=.78, p=.43, or for Rp–Neither objects, t(19)=.58, p=.57. Comparison between Rp+ and Nrp A’ scores, did not show significant within-participant facilitation, t(19)=.70, p=.49.
Next, between-participants RIF and facilitation was examined in a 5 (Item Type: Rp+, Rp–Colour, Rp–Shape, Rp–Both, and Rp–Neither) × 2 (Group: control vs. practice) mixed ANOVA with repeated measures on Item Type. The main effect of Item Type was significant, F(4, 152)=9.26, p<.001, ηp2 =.20, as was the main effect of Group, F(1, 38)=11.26, p=.002, ηp2 =.23. The interaction was marginally significant, F(4, 152)=2.01, p=.09, ηp2 =.05. Planned comparisons to examine between-participant RIF showed a significant difference between the practice and the control groups in A’ of Rp–Colour, Rp–Shape and Rp–Both objects [t(38)=2.80, p=.008, d=1.04; t(38)=2.41, d=.55, p=.02; and t(38)=2.91, p=.006, d=.21, respectively], but no significant RIF for Rp–Neither objects, t(38)=1.26, p=.21. There was no significant difference in A’ between the practice and control group for Rp+ items, t(38)=.09, p=.92, suggesting no significant between-participant facilitation.
Experiment 1 showed significant RIF for Rp–Shape and Rp–Colour objects, suggesting that these two object properties are encoded and independently drive competition effects in memory. Conclusions about independence are warranted here because Rp–Shape objects share only shape, but not colour, with the Rp+ objects and, thus, recognition-induced forgetting can only occur due to Rp+ and Rp– items having shape in common. Similarly, Rp–Colour objects share only colour, but not shape, and, therefore, forgetting in the Rp–Colour condition can only be due to Rp+ and Rp– items having colour in common.
The lack of facilitation for Rp+ objects is not without precedence (for review, see Storm & Levy, 2012). In retrieval practice tasks, forgetting of unpractised items (Rp– items) has often been obtained even when retrieval attempts fail completely (e.g., Storm, Bjork, Bjork, & Nestojko, 2006) and in the absence of facilitation for practised items (e.g., Gómez-Ariza, Fernandez & Bajo, 2012; Gómez-Ariza, Lechuga, Pelegrina, & Bajo, 2005; Maxcey & Bostic, 2015; Maxcey, Bostic, & Maldonado, 2016). Thus, the lack of facilitation for Rp+ objects in Experiment 1 does not detract from the current observation of recognition-induced forgetting for unpractised objects.
Based on prior findings, we initially expected that we might find significant RIF for Rp–Neither objects because they shared the same category as the practised objects (e.g., Maxcey & Woodman, 2014). However, there was no evidence of RIF for these objects, which may be due to their visual distinctiveness from the large number of similar objects in their category (i.e., they differed on multiple dimensions, such as shape, colour, texture and material; e.g., see the Rp–Neither example objects in Fig. 1b). The distinctiveness of the Rp–Neither objects may have encouraged participants to become aware of the differences among the exemplars within different categories and may have protected them from forgetting (e.g., Macrae & Roseveare, 2002; Smith & Hunt, 2000).
There was no within-participant RIF for Rp–Both objects. At first blush this lack of forgetting for Rp–Both objects might seem to question the sensitivity of the paradigm to detect the object features that it is supposed to measure. Does the paradigm fail to detect a feature (e.g., colour) when it is presented together with another feature (e.g., shape)?
The absence of RIF for Rp–Both objects may be explained by the model of distributed item representation in memory proposed by Anderson and colleagues (e.g., Anderson et al., 2000a, b; Anderson & Spellman, 1995). According to this model, the probability of an item being recollected is dependent on the overlap of features between target (i.e., Rp+) and competitor (i.e., Rp–) representations (see Fig. 2). With high feature overlap (e.g., Fig. 4, Panel A) the net activation of primed features in the competitor representation will result in the elimination of recognition-induced forgetting, or even facilitation for the competitor (in this case Rp–Both objects). Conversely, with moderate overlap (e.g., Fig. 4, Panel B) between target and competitor the net suppression of features in the competitor representation would be expected to outweigh the net facilitation from primed features shared with the Rp+ item (e.g., category), and thus yield significant recognition-induced forgetting. In Experiment 1, the feature overlap between Rp+ and Rp–Both objects can be considered high (i.e., the two object types shared object parts, part configuration, and part colour), which may be the reason for the lack of significant RIF. Therefore, for feature-based representations like the ones we are tapping into here (as evidenced by the presence of RIF for Rp–Colour and Rp–Shape objects), the lack of RIF in the Rp–Both condition may confirm the sensitivity of the paradigm to object features. This issue is revisited in Experiment 2.