diff --git a/R/vector_rsa_model.R b/R/vector_rsa_model.R
index c120e874..202cfb04 100644
--- a/R/vector_rsa_model.R
+++ b/R/vector_rsa_model.R
@@ -158,9 +158,23 @@ compute_trial_scores <- function(obj, X) {
   precomputed <- obj$design$model_mat
   dissimilarity_matrix <- precomputed$Dexpanded
   cross_block_data     <- precomputed$cross_block_data
-  
+
   # Convert X to matrix if needed
   X <- as.matrix(X)
+
+  # Basic dimensionality checks
+  if (nrow(X) != nrow(dissimilarity_matrix)) {
+    stop(
+      "nrow(X) (", nrow(X), ") must match nrow(precomputed$Dexpanded) (",
+      nrow(dissimilarity_matrix), ")"
+    )
+  }
+  if (length(obj$design$block) != nrow(X)) {
+    stop(
+      "length(obj$design$block) (", length(obj$design$block),
+      ") must match nrow(X) (", nrow(X), ")"
+    )
+  }
   
   # This function computes second-order similarity:
   #   second_order_similarity(distfun, X, D, block_var, rsa_simfun)
diff --git a/tests/testthat/test_compute_trial_scores.R b/tests/testthat/test_compute_trial_scores.R
new file mode 100644
index 00000000..f73cbeae
--- /dev/null
+++ b/tests/testthat/test_compute_trial_scores.R
@@ -0,0 +1,39 @@
+library(testthat)
+
+context("compute_trial_scores input validation")
+
+# Helper to set up a simple vector RSA model
+make_simple_model <- function(n) {
+  dset <- gen_sample_dataset(c(2,2,2), n, blocks = 3)
+  D <- as.matrix(dist(matrix(rnorm(n * n), n, n)))
+  labels <- paste0("L", seq_len(n))
+  rownames(D) <- labels
+  colnames(D) <- labels
+  rdes <- vector_rsa_design(D = D,
+                            labels = sample(labels, length(dset$design$block_var), replace = TRUE),
+                            block_var = dset$design$block_var)
+  vector_rsa_model(dset$dataset, rdes, distfun = cordist())
+}
+
+# nrow(X) vs nrow(precomputed$Dexpanded)
+
+test_that("compute_trial_scores errors when rows do not match Dexpanded", {
+  model <- make_simple_model(20)
+  X_wrong <- matrix(rnorm(25), 5, 5) # 5 rows, but Dexpanded is 20
+  expect_error(
+    compute_trial_scores(model, X_wrong),
+    "nrow\\(X\\).*nrow\\(precomputed\\$Dexpanded\\)"
+  )
+})
+
+# length(block) vs nrow(X)
+
+test_that("compute_trial_scores errors when block length mismatches", {
+  model <- make_simple_model(20)
+  X <- matrix(rnorm(100), 20, 5)
+  model$design$block <- model$design$block[-1]
+  expect_error(
+    compute_trial_scores(model, X),
+    "length\\(obj\\$design\\$block\\).*nrow\\(X\\)"
+  )
+})