Written and maintained by CASRAI Editorial Board
Published
On this page
Semantic priming is a quantitative psycholinguistic and cognitive phenomenon where prior exposure to a semantically related context stimulus (the prime) accelerates response latency and reduces error rates during the processing of a subsequent target stimulus compared to an unrelated baseline. Within quantitative language research, experimental psychologists and computational linguists measure semantic priming to map lexical memory architecture, evaluate spreading activation across conceptual networks, and isolate automatic retrieval operations from controlled attentional expectancies.
Definition and Theoretical Foundations
In quantitative psycholinguistics and experimental psychology, semantic priming refers to the facilitation of target word recognition produced by the prior presentation of a conceptually related prime word. When a participant is exposed to an initial stimulus such as doctor, their cognitive system processes a subsequent target word such as nurse significantly faster and with fewer categorization errors than if the prime had been an unrelated noun such as desk or cloud. This experimental latency reduction represents one of the most widely replicated behavioral effects in the quantitative cognitive sciences.
The foundational empirical demonstration of semantic priming was conducted by Meyer and Schvaneveldt in 1971 using a visual lexical decision task (LDT). In their paradigm, participants decided whether simultaneous or sequential strings of letters formed legitimate English words. Word pairs sharing a semantic or associative relationship exhibited response latencies that were 85 milliseconds faster than unrelated control pairs. This empirical finding challenged serial retrieval models and provided initial evidence that lexical retrieval operations are functionally interdependent rather than isolated serial lookups.
To account for these findings, Collins and Loftus (1975) introduced the spreading-activation theory of semantic memory. Under this architectural model, semantic memory is organized as an associative network where nodes represent concepts and connecting pathways represent semantic relations. The physical distance between nodes corresponds inversely to semantic similarity. When a prime word is perceived, its corresponding semantic node is activated. This excitation automatically spreads outward across intersecting network pathways, pre-activating adjacent nodes and temporarily lowering their activation thresholds. In modern language research within the discipline of linguistics, semantic priming paradigms serve as a primary empirical methodology for testing distributed representation models and evaluating lexical access dynamics.
Subsequent investigations by Neely (1977, 1991) established a critical theoretical distinction between automatic spreading activation and controlled, expectancy-based processing. Automatic spreading activation is fast-acting, capacity-free, and operates without conscious intention. In contrast, controlled processing involves conscious attentional expectancies and retrospective semantic matching, which require substantial cognitive capacity and take several hundred milliseconds to develop.
Mathematical Formulas and Quantitative Metrics
Quantifying semantic priming involves calculating mean difference scores across experimental conditions, standardizing these differences into parametric effect sizes, and fitting generalized linear mixed models to account for simultaneous participant and stimulus variability.
Response Time Priming Effect
The basic reaction time priming effect, denoted as PERT or ΔRT, is calculated as the arithmetic difference between the mean response time on unrelated trials and the mean response time on related trials:
PERT = Mean(RTunrelated) – Mean(RTrelated)
Where Mean(RTunrelated) represents the average response latency (in milliseconds) for correct trials where prime and target share no semantic relationship, and Mean(RTrelated) represents the average response latency for correct trials where prime and target share a documented semantic relationship. A positive priming score indicates facilitation, meaning that semantic relatedness speeded target recognition.
Facilitation and Inhibition Relative to a Neutral Baseline
To isolate the distinct cognitive mechanisms identified by Neely (1977), researchers frequently include a neutral baseline prime condition (such as a string of neutral characters like ###### or a neutral word like blank). This allows the partitioning of the overall priming effect into facilitation and inhibition components:
Facilitation = Mean(RTneutral) – Mean(RTrelated)
Inhibition = Mean(RTunrelated) – Mean(RTneutral)
Facilitation reflects the processing speedup gained from valid semantic pre-activation, whereas inhibition reflects the cognitive processing cost or attentional penalty incurred when an unrelated prime misdirects cognitive expectancy.
Error Rate Priming Effect
Priming also manifests in categorization accuracy, particularly in speeded lexical decision or semantic classification tasks. The error rate priming effect (PEER) is calculated as:
PEER = ErrorRateunrelated – ErrorRaterelated
A positive error rate priming effect confirms that semantic relatedness reduces classification errors alongside latency reductions, verifying that speed improvements do not stem from a speed-accuracy trade-off.
Effect Size: Paired Cohen’s d
Because semantic priming is typically evaluated within subjects, standardizing the effect requires the paired-samples standardized mean difference, known as Cohen’s dz:
dz = Mean(D) / sD
Where Mean(D) is the sample mean of the individual subject difference scores (Di = RTunrelated,i – RTrelated,i) and sD is the sample standard deviation of those paired differences. To compare results across independent studies or meta-analytic syntheses, researchers consult standardized effect size metrics such as Hedges’ gav.
Linear Mixed-Effects Model Formulation
While repeated-measures ANOVAs were historically common, contemporary quantitative psycholinguistics relies on Linear Mixed-Effects Models (LMMs) with crossed random effects for subjects and items (Baayen, Davidson, and Bates, 2008). The formal model equation is expressed as:
Yijk = β0 + β1 · Xijk + u0j + u1j · Xijk + w0k + w1k · Xijk + εijk
Where Yijk is the response latency for trial i from participant j responding to target item k; β0 is the fixed population intercept; β1 is the fixed effect slope of prime relatedness; u0j and u1j are random intercepts and slopes for participant j; w0k and w1k are random intercepts and slopes for target item k; and εijk is the residual trial-level error term.
Worked Example: Paired Reaction Time Calculation
To demonstrate the step-by-step arithmetic required to calculate semantic priming, consider a laboratory lexical decision experiment conducted with 10 participants. Each participant completed 80 trials (40 semantically related word pairs, 40 semantically unrelated word pairs) presented at a short stimulus onset asynchrony of 200 milliseconds. Below are the mean response times (in milliseconds) recorded for correct trials across both conditions.
| Participant ID | Unrelated RT (ms) | Related RT (ms) | Difference Di (ms) | Deviation (Di – Mean(D)) | Squared Deviation (Di – Mean(D))² |
|---|---|---|---|---|---|
| P01 | 685 | 652 | 33 | -3.0 | 9.0 |
| P02 | 714 | 668 | 46 | +10.0 | 100.0 |
| P03 | 650 | 626 | 24 | -12.0 | 144.0 |
| P04 | 692 | 671 | 21 | -15.0 | 225.0 |
| P05 | 738 | 685 | 53 | +17.0 | 289.0 |
| P06 | 660 | 638 | 22 | -14.0 | 196.0 |
| P07 | 722 | 672 | 50 | +14.0 | 196.0 |
| P08 | 680 | 655 | 25 | -11.0 | 121.0 |
| P09 | 705 | 651 | 54 | +18.0 | 324.0 |
| P10 | 674 | 642 | 32 | -4.0 | 16.0 |
Step 1: Compute Condition Means and Mean Priming Effect
First, calculate the sample means for the unrelated and related conditions:
Mean(RTunrelated) = (685 + 714 + 650 + 692 + 738 + 660 + 722 + 680 + 705 + 674) / 10 = 6920 / 10 = 692.0 ms
Mean(RTrelated) = (652 + 668 + 626 + 671 + 685 + 638 + 672 + 655 + 651 + 642) / 10 = 6560 / 10 = 656.0 ms
The mean semantic priming facilitation effect is:
Mean(D) = PERT = 692.0 ms – 656.0 ms = 36.0 ms
Step 2: Calculate Sum of Squared Deviations and Standard Deviation
Next, sum the squared deviations of the difference scores around their mean (Mean(D) = 36.0):
Σ (Di – Mean(D))² = 9 + 100 + 144 + 225 + 289 + 196 + 196 + 121 + 324 + 16 = 1620.0
With sample size N = 10, degrees of freedom are df = N – 1 = 9. The sample variance of the difference scores is:
sD² = 1620.0 / 9 = 180.0 ms²
Taking the square root yields the standard deviation of differences:
sD = √180.0 ≈ 13.416 ms
Step 3: Compute Standard Error and Test Statistic
Calculate the standard error of the mean difference (SED):
SED = sD / √N = 13.4164 / √10 = 13.4164 / 3.1623 ≈ 4.243 ms
Now compute the paired-samples t-statistic testing the null hypothesis of zero priming (H0: μD = 0):
t = Mean(D) / SED = 36.0 / 4.2426 ≈ 8.485
For df = 9, the two-tailed critical t-value at α = 0.05 is tcrit = 2.262. Because our obtained t = 8.485 exceeds 2.262, the observed priming facilitation is statistically significant with a corresponding p-value below .001.
Step 4: Determine Confidence Interval and Effect Size
Calculate the 95% confidence interval for the mean priming effect:
95% CI = Mean(D) ± (tcrit · SED) = 36.0 ± (2.262 · 4.2426) = 36.0 ± 9.60 = [26.40, 45.60] ms
Finally, compute the paired-samples standardized effect size Cohen’s dz:
dz = Mean(D) / sD = 36.0 / 13.4164 ≈ 2.683
This indicates an exceptionally robust experimental effect size, well above conventional benchmarks for behavioral latency studies.
Interpretation and Benchmark Values
Interpreting semantic priming magnitude requires evaluating task constraints, temporal presentation parameters, and lexical characteristics. In visual word recognition literature, priming effects typically range between 15 ms and 55 ms, though specific paradigms produce distinct empirical distributions.
The data table below compares empirical priming magnitudes, effect sizes, and operational parameters across seminal experimental publications. In addition to summary metrics, researchers evaluate effect sizes against the qualitative benchmarks shown in the following interpretation guide.
| Priming Effect (PERT) | Standardized dz | Statistical Interpretation | Typical Experimental Context |
|---|---|---|---|
| Under 10 ms | Under 0.20 | Negligible or Sub-threshold | Unassociated category coordinates at short SOA; masked priming with degraded primes. |
| 10 ms to 25 ms | 0.20 to 0.45 | Small to Moderate | Standard pronunciation or naming tasks; pure semantic category relations (e.g., horse-bear). |
| 25 ms to 45 ms | 0.45 to 0.80 | Moderate to Robust | Standard lexical decision tasks with associative pairs (e.g., doctor-nurse) at short SOAs (200-250 ms). |
| 45 ms to 70 ms | 0.80 to 1.20 | Large Facilitation | Long SOA paradigms (700-1200 ms) with high relatedness proportions engaging active expectancy. |
| Over 70 ms | Over 1.20 | Very Large or Compound | Compound associative and categorical relations; simultaneous pair presentation; degraded target stimuli. |
When interpreting priming effects, three structural factors dominate the observed latency differences:
- Task Demands: Lexical decision tasks consistently produce larger priming effects (25 to 45 ms) than pronunciation or naming tasks (10 to 20 ms). In LDT, participants can exploit semantic relatedness as an informative heuristic to decide that the target is a real word (retrospective semantic matching), whereas naming requires direct phonological transcoding where semantic matching provides less utility.
- Associative Strength vs. Pure Categorical Overlap: Pairs sharing both associative relations (frequent co-occurrence, e.g., cradle-baby) and categorical relations (taxonomic membership, e.g., robin-sparrow) produce the largest priming effects. Pure categorical pairs without high free-association frequency produce smaller but reliable priming effects (Lucas, 2000).
- Baseline Latency: Slower participant groups (such as older adults or clinical populations) systematically exhibit larger raw millisecond priming effects. Researchers apply z-score transformations or proportional priming calculations (PEprop = PERT / RTunrelated) to control for general cognitive slowing.
Methodological Controls, Assumptions, and Common Mistakes
Rigorous semantic priming research demands strict control over experimental timing, stimulus properties, and list context to prevent confounding variables from contaminating the observed effects.
1. Stimulus Onset Asynchrony (SOA)
Stimulus Onset Asynchrony is the time interval elapsed between the visual onset of the prime and the visual onset of the target. Manipulating SOA is the primary empirical tool for dissociating automatic from strategic processes:
- Short SOA (250 ms or less): Prevents participants from developing conscious expectancies before the target appears. Priming observed at short SOAs reflects rapid, capacity-free automatic spreading activation within the lexicon.
- Long SOA (500 ms or more, typically 700 to 1200 ms): Provides sufficient time for participants to consciously generate an expectancy set based on the prime. Long SOAs capture controlled attentional processes, producing substantial facilitation for expected targets and inhibition for unexpected targets.
2. Relatedness Proportion (RP)
The Relatedness Proportion is the ratio of related word-word pairs to the total number of word-word pairs in the experimental list. When RP is high (e.g., 0.70 to 0.80), participants quickly deduce that primes reliably predict target identities, inducing conscious expectancy strategies that inflate priming effects. When RP is low (e.g., 0.20 to 0.33), participants abandon expectancy strategies, allowing researchers to isolate automatic activation mechanisms.
3. Nonword Ratio and Retrospective Semantic Matching
In lexical decision tasks, the composition of the nonword filler trials can introduce severe strategic artifacts. The Nonword Ratio represents the probability that a target is a nonword given that the prime and target are unrelated. If all unrelated trials consist of words while nonwords are paired only with neutral primes, participants can use the presence of semantic relatedness as an immediate shortcut to press the ‘Word’ button. To prevent retrospective semantic matching heuristics, experimenters must ensure that nonwords are matched in orthographic legality, phonotactic plausibility, and pseudoword length.
4. Stimulus Matching Across Conditions
A pervasive flaw in older psycholinguistic research was failing to match target words across related and unrelated conditions on baseline lexical characteristics. Prime and target sets must be counterbalanced across participant lists using Latin square designs so that identical target words appear in both related and unrelated conditions. Critical lexical dimensions that must be strictly matched include:
- Word Frequency: Sourced from contemporary subtitle corpora such as SUBTLEX-US (Brysbaert and New, 2009). Target frequency correlates negatively with response latency.
- Orthographic Neighborhood Density: The number of words that can be formed by altering a single letter (Coltheart’s N).
- Length and Bigram Frequency: Character count, syllable count, and printed letter co-occurrence statistics.
- Concreteness and Imageability: Semantic feature ratings obtained from standardized psycholinguistic norms.
5. Data Trimming and Distributional Assumptions
Raw reaction times do not follow Gaussian normal distributions; they are heavily right-skewed with long tails. Common analytical errors include applying standard parametric t-tests to untransformed raw latency data containing extreme outliers. Best practices require:
- Removing anticipatory errors (responses under 200 ms) and gross attentional lapses (responses exceeding 2000 ms).
- Applying recursive outlier trimming (e.g., excluding trials exceeding 2.5 or 3 standard deviations from each participant’s cell mean) or utilizing non-parametric median metrics.
- Applying mathematical transformations such as natural logarithm or reciprocal latency prior to linear modeling to satisfy normality and homoscedasticity assumptions. For categorical error rates across conditions, researchers employ non-parametric evaluations such as the chi-square test or binomial generalized linear mixed models.
How to Report Semantic Priming in APA 7th Style
Reporting semantic priming results under American Psychological Association (APA 7th edition) guidelines requires complete statistical disclosure, including descriptive statistics (means and standard deviations for all conditions), exact difference scores, 95% confidence intervals, precise test statistics with degrees of freedom, exact p-values, and standardized effect sizes.
APA Reporting Template: Paired-Samples t-Test
When reporting a traditional repeated-measures latency analysis, use the following format:
A paired-samples t-test was conducted to assess the effect of semantic relatedness on visual word recognition latencies. Target words preceded by semantically related primes (M = 656.0 ms, SD = 18.04) were recognized significantly faster than target words preceded by semantically unrelated primes (M = 692.0 ms, SD = 27.84), revealing a significant semantic priming facilitation effect of 36.0 ms, 95% CI [26.40, 45.60], t(9) = 8.49, p < .001, Cohen's dz = 2.68. Error rates in the related prime condition (M = 2.1%, SD = 1.1) were also significantly lower than in the unrelated prime condition (M = 4.8%, SD = 1.6), t(9) = 4.31, p = .002, dz = 1.36, confirming that faster response times did not reflect a speed-accuracy trade-off.
APA Reporting Template: Linear Mixed-Effects Model
When reporting modern trial-level linear mixed-effects models, report fixed effect estimates (b), standard errors, t-values, and model convergence specifications:
Trial-level reaction times for correct trials were analyzed using a linear mixed-effects model fitted with the
lme4package in R. The model included prime condition (contrast-coded: -0.5 = related, +0.5 = unrelated) as a fixed factor, along with crossed random intercepts and slopes for both participants and target items. Prime relatedness exerted a significant fixed effect on response latency, b = -36.14 ms, SE = 6.82, t = -5.30, p < .001, indicating that semantic priming facilitated recognition by approximately 36 ms after accounting for baseline subject speed and item-level word frequency.
How to Calculate Semantic Priming in Excel, R, and Python
Below are reproducible, tested code snippets for calculating semantic priming effect metrics, executing statistical hypothesis testing, and determining effect sizes across common analytical platforms.
1. Microsoft Excel
Assuming participant mean response times are arranged in columns: Column A (Participant ID, rows 2 to 11), Column B (Unrelated RT in ms), and Column C (Related RT in ms):
- Difference Score (Cell D2):
=B2-C2(copy down through row 11) - Mean Priming Effect:
=AVERAGE(D2:D11) - Standard Deviation of Differences (sD):
=STDEV.S(D2:D11) - Standard Error of the Difference (SED):
=STDEV.S(D2:D11)/SQRT(COUNT(D2:D11)) - Paired t-Statistic:
=AVERAGE(D2:D11)/(STDEV.S(D2:D11)/SQRT(COUNT(D2:D11))) - Two-Tailed p-Value:
=T.TEST(B2:B11, C2:C11, 2, 1) - Paired Cohen’s dz:
=AVERAGE(D2:D11)/STDEV.S(D2:D11) - 95% CI Lower Bound:
=AVERAGE(D2:D11)-T.INV.2T(0.05, COUNT(D2:D11)-1)*(STDEV.S(D2:D11)/SQRT(COUNT(D2:D11))) - 95% CI Upper Bound:
=AVERAGE(D2:D11)+T.INV.2T(0.05, COUNT(D2:D11)-1)*(STDEV.S(D2:D11)/SQRT(COUNT(D2:D11)))
2. R Code Snippet
# Load required libraries
library(lme4)
library(lmerTest)
# Participant-level paired t-test and effect size
unrelated <- c(685, 714, 650, 692, 738, 660, 722, 680, 705, 674)
related <- c(652, 668, 626, 671, 685, 638, 672, 655, 651, 642)
diff <- unrelated - related
# Statistical t-test
t_res <- t.test(unrelated, related, paired = TRUE)
pe_mean <- mean(diff)
s_d <- sd(diff)
cohen_dz <- pe_mean / s_d
cat(sprintf("Priming Effect: %.2f msn", pe_mean))
cat(sprintf("t(%d) = %.3f, p = %.4e, Cohen's dz = %.3fn",
t_res$parameter, t_res$statistic, t_res$p.value, cohen_dz))
# Trial-level linear mixed-effects model (Baayen et al., 2008)
# Assuming a data frame 'trials' with columns: rt, prime_cond, subject_id, item_id
# model <- lmer(rt ~ prime_cond + (1 + prime_cond | subject_id) +
# (1 + prime_cond | item_id), data = trials)
# summary(model)
3. Python Code Snippet
import math
import statistics
# Participant latency vectors
unrelated = [685, 714, 650, 692, 738, 660, 722, 680, 705, 674]
related = [652, 668, 626, 671, 685, 638, 672, 655, 651, 642]
# Calculate difference scores
diff = [u - r for u, r in zip(unrelated, related)]
n = len(diff)
mean_diff = statistics.mean(diff)
s_d = statistics.stdev(diff)
se_d = s_d / math.sqrt(n)
t_stat = mean_diff / se_d
cohen_dz = mean_diff / s_d
# 95% Confidence Interval for df=9 (critical t = 2.262)
t_crit = 2.262
ci_lower = mean_diff - (t_crit * se_d)
ci_upper = mean_diff + (t_crit * se_d)
print(f"Mean Priming Effect: {mean_diff:.1f} ms")
print(f"SD of Differences: {s_d:.3f} ms")
print(f"Standard Error: {se_d:.3f} ms")
print(f"t({n-1}) = {t_stat:.3f}, Cohen's dz = {cohen_dz:.3f}")
print(f"95% CI: [{ci_lower:.2f}, {ci_upper:.2f}] ms")
# Trial-level modeling using statsmodels:
# import statsmodels.formula.api as smf
# lmm = smf.mixedlm("rt ~ prime_condition", data=df, groups=df["subject_id"]).fit()
# print(lmm.summary())
Frequently asked questions
What is the difference between semantic priming and associative priming?
Semantic priming occurs between words that share conceptual, taxonomic, or feature overlap (such as horse and bear, which are both four-legged mammals), whereas associative priming occurs between words that frequently co-occur in language usage and free-association norms (such as cradle and baby). While many word pairs are both semantically and associatively related (e.g., doctor and nurse), Lucas (2000) demonstrated that pure semantic relations produce reliable priming even in the absence of associative co-occurrence.
How does Stimulus Onset Asynchrony (SOA) influence semantic priming?
Stimulus Onset Asynchrony (SOA) dictates whether automatic or strategic mechanisms generate the observed priming effect. At short SOAs (under 250 milliseconds), participants do not have sufficient time to consciously process the prime or form expectations, isolating automatic spreading activation. At long SOAs (500 to 1200 milliseconds), participants consciously use the prime to predict likely target words, engaging controlled attentional expectancies and retrospective semantic matching.
What is the difference between facilitation and inhibition in priming experiments?
Facilitation is the response time acceleration observed for semantically related targets relative to a neutral prime baseline (e.g., ######), reflecting the benefit of pre-activated semantic representations. Inhibition is the response time deceleration observed for unrelated targets relative to that same neutral baseline, reflecting the cognitive cost of redirecting attention away from an unexpected or misdirected concept.
Why are linear mixed-effects models preferred over repeated-measures ANOVA for semantic priming?
Traditional repeated-measures ANOVAs require averaging across items (by-subject analysis, F1) or averaging across subjects (by-item analysis, F2), which creates a language-as-fixed-effect fallacy and struggles with missing trial data. Linear mixed-effects models (Baayen, Davidson, and Bates, 2008) analyze trial-level responses while simultaneously fitting crossed random intercepts and random slopes for both subjects and items, properly partitioning variance without data aggregation.
How does the relatedness proportion affect automatic versus strategic priming?
The relatedness proportion is the percentage of word pairs in an experiment that are semantically related. When the proportion is high (e.g., 75%), participants quickly recognize that primes predict targets, actively generating expectancy lists and increasing priming magnitudes. When the proportion is low (e.g., 20%), participants abandon expectancy strategies, ensuring that measured effects reflect automatic spreading activation rather than conscious problem-solving.
Can semantic priming occur without conscious awareness of the prime?
Yes, semantic priming occurs under masked conditions where the prime is presented for a very brief duration (typically 30 to 50 milliseconds) and surrounded by visual backward and forward masks (e.g., strings of hash marks). Under these subliminal conditions, participants cannot consciously report or identify the prime word, yet they still exhibit significant reaction time facilitation to related targets, demonstrating automatic lexical activation.
Sources
- Baayen, R. H., Davidson, D. J., and Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59(4), 390–412. https://doi.org/10.1016/j.jml.2007.12.005 (Accessed 2026-11-14)
- Brysbaert, M., and New, B. (2009). Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41(4), 977–990. https://doi.org/10.3758/BRM.41.4.977 (Accessed 2026-11-14)
- Collins, A. M., and Loftus, E. F. (1975). A spreading-activation theory of semantic memory. Psychological Review, 82(6), 407–428. https://doi.org/10.1037/0033-295X.82.6.407 (Accessed 2026-11-14)
- Hutchison, K. A., Balota, D. A., Neely, J. H., Cortese, M. J., Cohen-Shikora, E. R., Tse, C.-S., Yap, M. J., Bengson, J. J., Niemeyer, D., and Buchanan, E. (2013). The Semantic Priming Project. Behavior Research Methods, 45(4), 1099–1114. https://doi.org/10.3758/s13428-012-0304-z (Accessed 2026-11-14)
- Lucas, M. (2000). Semantic priming without association: A meta-analytic review. Psychonomic Bulletin and Review, 7(4), 618–630. https://doi.org/10.3758/BF03212999 (Accessed 2026-11-14)
- Meyer, D. E., and Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology, 90(2), 227–234. https://doi.org/10.1037/h0031564 (Accessed 2026-11-14)
- Neely, J. H. (1977). Semantic priming and retrieval from lexical memory: Roles of inhibitionless spreading activation and limited-capacity attention. Journal of Experimental Psychology: General, 106(3), 226–254. https://doi.org/10.1037/0096-3445.106.3.226 (Accessed 2026-11-14)
- Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner and G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264–336). Lawrence Erlbaum Associates. (Accessed 2026-11-14)
Data: Semantic Priming Benchmarks Across Paradigms
Empirical comparison of semantic priming latency effects, effect sizes, and experimental parameters across landmark psycholinguistic publications compiled from peer-reviewed experimental literature.
Period covered: 1971/2013. Coverage: United States.
| Study and Experimental Condition | Task Type | Stimulus Onset Asynchrony (ms) | Relatedness Proportion (proportion) | Unrelated Mean RT (ms) | Related Mean RT (ms) | Priming Effect (ms) | Effect Size (Cohen d) | Primary Cognitive Mechanism |
|---|---|---|---|---|---|---|---|---|
| Meyer & Schvaneveldt (1971) Simultaneous LDT | Lexical Decision | 0 | 0.5 | 940 | 855 | 85 | 0.72 | Automatic Retrieval Dependence |
| Neely (1977) Nonshift Short SOA | Lexical Decision | 250 | 0.5 | 686 | 648 | 38 | 0.48 | Automatic Spreading Activation |
| Neely (1977) Shift Expected Long SOA | Lexical Decision | 700 | 0.67 | 682 | 631 | 51 | 0.64 | Attentional Expectancy |
| Hutchison et al. (2013) SPP LDT Short SOA | Lexical Decision | 200 | 0.5 | 651 | 624 | 27 | 0.36 | Automatic Spreading Activation |
| Hutchison et al. (2013) SPP LDT Long SOA | Lexical Decision | 1200 | 0.5 | 644 | 601 | 43 | 0.58 | Semantic Matching and Expectancy |
| Hutchison et al. (2013) SPP Naming Short SOA | Pronunciation / Naming | 200 | 0.5 | 563 | 551 | 12 | 0.22 | Automatic Spreading Activation |
| Hutchison et al. (2013) SPP Naming Long SOA | Pronunciation / Naming | 1200 | 0.5 | 559 | 544 | 15 | 0.28 | Articulatory Expectancy |
| Lucas (2000) Unassociated Taxonomic Pairs | Lexical Decision | 250 | 0.5 | 635 | 612 | 23 | 0.31 | Pure Conceptual Feature Overlap |
Sources
- Meyer & Schvaneveldt (1971) J Exp Psychol (accessed 2026-11-14)
- Neely (1977) J Exp Psychol Gen (accessed 2026-11-14)
- Hutchison et al. (2013) Behav Res Methods (accessed 2026-11-14)
- Lucas (2000) Psychon Bull Rev (accessed 2026-11-14)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







