diff --git a/NEWS.md b/NEWS.md
index 23e8d5c87..4032e03a7 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -16,6 +16,8 @@
 
 1. `nafill()`, `setnafill()` extended to work on logical vectors (part of [#3992](https://github.com/Rdatatable/data.table/issues/3992)). Thanks @jangorecki for the request and @MichaelChirico for the PR.
 
+2. Joins (`y[x, on=]` or `merge(x, y, ...)`) now display join statistics with `options(datatable.verbose=TRUE)`, showing row counts, matched rows, and join columns used, [#4677](https://github.com/Rdatatable/data.table/issues/4677). Thanks @thorek1 and @grantmcdermott for the suggestion and @ben-schwen for the implementation.
+
 ### Notes
 
 1. {data.table} now depends on R 3.5.0 (2018).
diff --git a/R/bmerge.R b/R/bmerge.R
index 3c903ae35..00113adc3 100644
--- a/R/bmerge.R
+++ b/R/bmerge.R
@@ -25,7 +25,7 @@ coerce_col = function(dt, col, from_type, to_type, from_name, to_name, from_deta
   set(dt, j=col, value=cast_with_attrs(dt[[col]], cast_fun))
 }
 
-bmerge = function(i, x, icols, xcols, roll, rollends, nomatch, mult, ops, verbose)
+bmerge = function(i, x, icols, xcols, roll, rollends, nomatch, mult, ops, verbose, notjoin=FALSE)
 {
   if (roll != 0.0 && length(icols)) {
     last_x_idx = tail(xcols, 1L)
@@ -224,6 +224,41 @@ bmerge = function(i, x, icols, xcols, roll, rollends, nomatch, mult, ops, verbos
   if (verbose) {catf("bmerge done in %s\n",timetaken(last.started.at)); flush.console()}
   # TO DO: xo could be moved inside Cbmerge
 
+  # join statistics
+  if (verbose) {
+    nrow_x = nrow(x)
+    nrow_i = nrow(i)
+    inner_join = is.null(nomatch) || identical(nomatch, 0L)
+    idx = if (inner_join) ans$starts != 0L else !is.na(ans$starts)
+    matched_i = sum(idx)
+
+    if (notjoin) {
+      # Anti-join: count rows in x that were NOT matched
+      result_rows = if (matched_i > 0L) nrow_x - length(unique(ans$starts[idx])) else nrow_x
+    } else if (inner_join) {
+      # Inner join: sum lengths for matched rows only
+      result_rows = if (matched_i > 0L) sum(ans$lens[idx]) else 0L
+    } else {
+      # Left join: sum all lengths (includes NAs for unmatched)
+      result_rows = sum(ans$lens)
+    }
+
+    op_symbols = if (length(ops)) c("==", "<=", "<", ">=", ">")[ops] else strrep("==", length(icols))
+    join_str = toString(sprintf("%s %s %s", names(x)[xcols], op_symbols, names(i)[icols]))
+    num_width = max(vapply_1i(list(nrow_x, nrow_i, matched_i, result_rows, join_str), nchar))
+    # nchar("rows in x:    ") == 14L
+    separator = strrep("-", 14L + num_width)
+
+    catf("Join summary:\n")
+    catf("  rows in x:    %*d\n", num_width, nrow_x)
+    catf("  rows in i:    %*d\n", num_width, nrow_i)
+    catf("  matched rows: %*d\n", num_width, matched_i)
+    catf("  join columns: %s\n", join_str)
+    catf("  %s\n", separator)
+    catf("  result rows:  %*d\n", num_width, result_rows)
+    flush.console()
+  }
+
   ans$xo = xo  # for further use by [.data.table
   ans
 }
diff --git a/R/data.table.R b/R/data.table.R
index 27c985e44..e2ac9b30b 100644
--- a/R/data.table.R
+++ b/R/data.table.R
@@ -518,7 +518,7 @@ replace_dot_alias = function(e) {
         setattr(i, 'sorted', names(i)) # since 'x' has key set, this'll always be sorted
       }
       i = .shallow(i, retain.key = TRUE)
-      ans = bmerge(i, x, leftcols, rightcols, roll, rollends, nomatch, mult, ops, verbose=verbose)
+      ans = bmerge(i, x, leftcols, rightcols, roll, rollends, nomatch, mult, ops, verbose=verbose, notjoin=notjoin)
       if (mult == "error") mult = "all" ## error should have been raised inside bmerge() call above already, if it wasn't continue as mult="all"
       xo = ans$xo ## to make it available for further use.
       # temp fix for issue spotted by Jan, test #1653.1. TODO: avoid this
diff --git a/inst/tests/tests.Rraw b/inst/tests/tests.Rraw
index fcf78e9f3..605e6083a 100644
--- a/inst/tests/tests.Rraw
+++ b/inst/tests/tests.Rraw
@@ -21978,3 +21978,22 @@ local({
   test(2357.1, fread(f), DT)
   test(2357.2, fread(paste0("file://", f)), DT)
 })
+
+# join statistics #4677
+x = data.table(A = 1:5, B = 6:10)
+y = data.table(A = c(1L, 1L, 4L), C = LETTERS[c(1L, 2L, 4L)])
+test(2358.1, nrow(x[y, on = "A", verbose=TRUE]), 3L, output="result rows: \\s+3\n")
+test(2358.2, nrow(y[x, on = "A", verbose=TRUE]), 6L, output="result rows: \\s+6\n")
+test(2358.3, nrow(y[x, on = "A", nomatch=NULL, verbose=TRUE]), 3L, output="result rows: \\s+3\n")
+test(2358.4, nrow(x[!y, on = "A", verbose=TRUE]), 3L, output="result rows: \\s+3\n")
+test(2358.5, nrow(y[y, on = "A", allow.cartesian=TRUE, verbose=TRUE]), 5L, output="result rows: \\s+5\n")
+test(2358.6, options=c(datatable.verbose=TRUE), nrow(merge(x, y, by="A")), 3L, output="result rows: \\s+3\n")
+x = data.table(id = c("A", "A", "A", "B", "B"), date = as.IDate(c("2010-01-01", "2012-01-01", "2014-01-01", "2010-01-01", "2012-01-01")))
+y = data.table(id = c("A", "B"), date = as.IDate(c("2013-01-01", "2013-01-01")))
+test(2358.7, nrow(x[y, on = .(id, date <= date), verbose=TRUE]), 4L, output="join columns: id == id, date <= date.*result rows: \\s+4\n")
+x = data.table(A = integer(0))
+y = data.table(A = 1:3)
+test(2358.8, nrow(x[y, on="A", verbose=TRUE]), 3L, output="result rows: \\s+3\n")
+x = data.table(A = 1:3)
+y = data.table(A = 4:6)
+test(2358.9, nrow(x[y, on="A", nomatch=NULL, verbose=TRUE]), 0L, output="matched rows: \\s+0\n")