Our need for associative coherence

The study investigates whether associative coherence—regular, predictable relations between items based on co-occurrence in everyday contexts—affects low-level visual perception, not just high-level cognition. While associations aid prediction, planning, memory encoding/retrieval, and recognition, their influence on early perceptual processes is less explored. The authors hypothesize that when associative coherence is absent, people persist in searching for links, imposing a lingering cognitive load that degrades subsequent perceptual performance. Using pairs of associated (e.g., monkey-banana) versus non-associated (e.g., grapes-flag) images to induce associative or non-associative processing, they test impacts on three perceptual tasks: contrast sensitivity, global vs. local (hierarchical) perception, and critical flicker fusion. A baseline control helps distinguish whether effects are due to associative priming or to a sustained search for coherence.

Prior work shows associative thinking relates to creativity, problem-solving, and prediction, and that associations are intrinsically attractive. Contextual associations facilitate scene and object recognition and memory processes. There is evidence for links between associativity and mood, and between positive mood and global perceptual bias, suggesting a possible chain from associativity to mood to perception. Global precedence in visual processing (global features before local) is well established, and low spatial frequencies can trigger top-down predictions that facilitate recognition. Cognitive load diminishes global precedence, indicating resource limitations can alter hierarchical processing. Spreading activation theories posit automatic activation of associated memory nodes, influencing processing fluency. Studies also show associative knowledge can guide attention and that working memory load and attentional resources modulate perceptual sensitivity (including CFF), reaction times, eye movements, and accuracy. Collectively, the literature supports that associativity, attention, mood, and cognitive load shape perception, motivating tests of whether absent associative coherence impairs low-level visual processing.

Overall design: Across three experiments, participants viewed pairs of object images that were either associatively related or non-associated, then performed a visual perceptual task. A baseline block (no images) was included in Experiments 1 and 2. Associative strength was based on Hebrew word association norms (Rubinsten et al., 2005), converted to pictures. Image luminance/contrast were equated; images were neutral, cut from original backgrounds, and presented on white background. An independent sample (n=20) rated image-pair relatedness (0–10); only pairs ≥8.5 (associated; mean 8.9) or ≤1.5 (non-associated; mean 1.2) were used; mid-range pairs were excluded. Catch trials with identical images ensured attention. Stimuli were presented on a 1920×1080 display; head stabilized at 55 cm. Block order randomized/counterbalanced; within-block trial order randomized. Ethics approval obtained; informed consent provided; different participants per experiment. Experiment 1 (Contrast Sensitivity): Participants: 41 students (24 females; 19–32 years; mean 24.17); one excluded for catch-trial errors. Stimuli: 310 neutral object images forming 148 pairs + 14 singles. Target: Gabor patch (2 cpd; Gaussian σ=1°; slanted ±45°). Procedure: Practice with 8 mixed pairs. Three blocks: baseline (Gabor only), associated, non-associated. Each block: 70 staircase trials; associated/non-associated blocks included 7 catch trials (77 total). Trial sequence (assoc/non-assoc): fixation (500 ms), image1 (300 ms), fixation (250 ms), image2 (300 ms), fixation (250 ms), Gabor (150 ms), response. Task: indicate Gabor orientation via Left/Right arrow; in catch trials, press Down Arrow ignoring Gabor. Threshold estimation: QUEST adaptive staircase (75% correct), yielding log10 contrast threshold; RTs recorded for correct trials; participants with >50% catch errors excluded. Experiment 2 (Global vs. Local perception): Participants: 40 students (27 females; 18–32 years; mean 24.05). Stimuli: 594 neutral object images forming 276 pairs + 42 singles. Target: Navon stimuli with global letters (H/S) composed of local letters (h/s); both consistent and inconsistent sets; global size 3×2°; local 0.4×0.35°. Each letter displayed 54 times per block. Procedure: Practice with 12 mixed pairs. Three blocks: baseline (Navon only), associated, non-associated. Each block: 216 experimental trials; associated/non-associated blocks had 21 catch trials (237 total). Trial sequence (assoc/non-assoc): fixation (500 ms), image1 (300 ms), fixation (250 ms), image2 (300 ms), fixation (250 ms), Navon (40 ms), response. Task: report which letter (H/S) was perceived first (global or local) via keypress; catch trials required pressing Space ignoring the Navon. No feedback; RT recorded; exclusion if >50% catch errors. Experiment 3 (Critical Flicker Fusion): Participants: 40 students (25 females; 18–33 years; mean 23.40). Stimuli: 520 neutral object images forming 248 pairs + 24 singles. Target: central white LED (5 mm, 20 mA) controlled by Arduino; flickering frequency adaptively varied. Procedure: To minimize discomfort, only associated and non-associated blocks (no baseline), each with 132 trials: 60 experimental, 60 fillers (random frequencies to decorrelate response habituation), 12 catch trials. Practice with 8 mixed pairs. Trial sequence: fixation (500 ms), image1 (300 ms), fixation (250 ms), image2 (300 ms), fixation (250 ms), flickering/continuous LED (100 ms), response (yes=continuous, no=flickering). FAST adaptive method estimated CFF threshold (0–60 Hz logistic function; slope 0.1–1). After a “continuous” response, frequency decreased; after “flickering”, frequency increased. Final threshold from 60 experimental trials per block. No feedback; RT recorded; exclusion if >50% catch errors.

Experiment 1 (Contrast Sensitivity): Repeated-measures ANOVA across baseline, associated, non-associated blocks showed a main effect on CS thresholds (baseline=0.00903 log intensity; associated=0.01090; non-associated=0.01494; F(2,78)=23.019, p<0.001, η²=0.371, BF10=1.628×10^6). Pairwise: baseline vs non-associated p<0.001, Cohen’s d=1.013, 95% CI [0.004, 0.008], BF10=106,781.1; associated vs non-associated p<0.001, d=0.782, CI [0.002, 0.006], BF10=1,403.42; baseline vs associated ns (p=0.153, d=0.318, CI [−0.0042, 0.00045], BF10=1.054). RT longer in non-associated (868.19 ms) than associated (850.73 ms), t(39)=2.134, p<0.05, d=0.337, CI [0.911, 34.023], BF10=1.302. Block order did not affect results. Experiment 2 (Global vs. Local): ANOVA showed a main effect on proportion choosing global first (baseline=79.14%; associated=80.88%; non-associated=69.86%; F(2,78)=7.811, p<0.001, η²=0.167, observed power=0.944, BF10≈34.92). Pairwise: baseline vs non-associated p<0.05, d=0.398, CI [0.0655, 18.499], BF10=2.742; associated vs non-associated p<0.005, d=0.547, CI [3.046, 18.990], BF10=23.674; baseline vs associated ns (p=1, d=0.153, CI [−6.234, 2.762], BF10=0.263). RT for global responses longer in non-associated (500.50 ms) than associated (447.13 ms), t(39)=2.234, p<0.05, d=0.535, CI [5.037, 101.70], BF10=1.564. Block order did not affect results. Experiment 3 (CFF): Paired t-test showed lower mean CFF threshold in non-associated (47.65 Hz) than associated (50.04 Hz), t(39)=−2.390, p<0.05, d=0.378, CI [0.369, 4.428], BF10=2.114. RTs did not differ (associated=343.85 ms; non-associated=356.90 ms; t(39)=0.746, p=0.459, d=0.118, CI [−22.32, 48.43], BF10=0.221). Block order did not affect results. Across experiments, lack of associative coherence consistently degraded subsequent perceptual performance (higher CS thresholds, reduced global precedence, lower CFF thresholds) and often slowed RTs, while associative and baseline conditions were comparable.

Findings support the hypothesis that absence of associative coherence imposes a lingering cognitive load due to a continued search for a link between unrelated stimuli, which then taxes resources needed for subsequent low-level perception. The equivalence of associative and baseline conditions argues against simple associative priming or general attentional facilitation as the primary driver; instead, non-associative pairs likely trigger sustained processing that reduces available resources. This aligns with spreading activation accounts: associated pairs yield a match between anticipated and presented items, enabling disengagement and task readiness, whereas non-associated pairs violate expectations, prolonging search for coherence. Alternative accounts, such as varying attentional allocation by associations, are considered, but the lack of associative-better-than-baseline effects across Experiments 1–2 weakens pure attentional facilitation explanations. The resource-limitation framework is consistent with prior evidence that cognitive load diminishes global precedence and reduces temporal discrimination (CFF). Working memory capacity and engagement may mediate these effects, potentially measurable via indices like pupil dilation. Overall, results indicate that associative structure in preceding context can modulate even early perceptual thresholds by altering cognitive load.

The study demonstrates that when associative coherence is lacking, people appear to persist in seeking coherence, imposing cognitive load that degrades performance on three low-level visual tasks: contrast sensitivity, hierarchical global precedence, and critical flicker fusion. The comparable performance between associative and baseline conditions suggests impairment arises from non-associative contexts rather than facilitation by associative priming. These findings highlight non-physical, context-driven influences—specifically associative coherence—on perception. Future research should elucidate the underlying mechanisms (e.g., direct measures of working memory load, pupil dilation), examine neural correlates of sustained search for coherence, test generalization across modalities and tasks, and explore boundary conditions (e.g., strength of associations, timing, mood interactions).

• Induction strength: It is difficult to claim that brief image pairs alone fully shift thought patterns; the manipulation should be viewed as a relative induction of associative versus non-associative processing.
• Baseline omission in Experiment 3: Due to discomfort concerns, CFF included only associated and non-associated blocks, limiting direct comparison to baseline.
• Potential RT constraints in CFF: The 100 ms fixed stimulus duration may have limited RT differences, as participants could not respond until offset.
• Generalizability: Samples consisted of university students with normal vision; broader populations and modalities were not tested.
• Mood and other mediators were not measured; although discussed theoretically, their contribution was not directly assessed.