diff --git a/src/04-modality-specific-files/05-task-events.md b/src/04-modality-specific-files/05-task-events.md
index 407f9ebc86..5b76a27943 100644
--- a/src/04-modality-specific-files/05-task-events.md
+++ b/src/04-modality-specific-files/05-task-events.md
@@ -52,7 +52,7 @@ and a guide for using macros can be found at
    }
 ) }}
 
-<sup>5</sup> Note for MRI data:
+Note for MRI data:
 If any acquired scans have been discarded before forming the imaging data file,
 ensure that an `onset` of 0 corresponds to the time the first image was stored.
 For example in case there is an in scanner training phase that
@@ -60,8 +60,14 @@ begins before the scanning sequence has started events from this sequence should
 have negative onset time counting down to the beginning of the acquisition of
 the first volume.
 
+Note regarding the precision of numeric metadata:
+It is RECOMMENDENDED that dataset curators specify numeric metadata like `onset` and
+`duration` with as much decimal precision as is reasonable in the context of the experiment.
+For example in an EEG experiment with devices operating at 1000 Hz sampling frequency,
+dataset curators SHOULD specify **at least** 3 figures after the decimal point.
+
 An arbitrary number of additional columns can be added. Those allow describing
-other properties of events that could be later referred in modelling and
+other properties of events that could be later referenced in modelling and
 hypothesis extensions of BIDS.
 Note that the `trial_type` and any additional columns in a TSV file
 SHOULD be documented in an accompanying JSON sidecar file.
diff --git a/src/schema/objects/columns.yaml b/src/schema/objects/columns.yaml
index b4cf79852b..60dda15844 100644
--- a/src/schema/objects/columns.yaml
+++ b/src/schema/objects/columns.yaml
@@ -347,6 +347,9 @@ sample:
   description: |
     Onset of the event according to the sampling scheme of the recorded modality
     (that is, referring to the raw data file that the `events.tsv` file accompanies).
+    When there are several sampling schemes present in the raw data file (as can be
+    the case for example for `.edf` files), this column is ambiguous and
+    SHOULD NOT be used.
   type: integer
 sample_id:
   name: sample_id