The aim of this study was to determine how many trials are required to assess VV perception after stroke to better adapt assessment to patient abilities and at the same time maintain good inter-trial reliability of the measure. Surprisingly, our study is the first to address the question of the number of trials sufficient for VV measurement.
To the best of our knowledge, among the myriad studies of VV perception using different numbers of trials, no studies have ever justified the choice of this number. Our study showed that the number of trials needed to reach high inter-trial reliability was a minimum of six for patients with contralesional and ipsilesional biases. For this number of trials, the SEM was slightly higher than 1° for the contralesional bias subgroup. For patients with normal VV perception, it seems that more than 10 trials are required, but the SEM (1.13°) with this number was as good as for the other subgroups. As compared with 10 trials, 6 and 8 trials allowed for a correct identification of alterations in VV perception in most patients, with less than 5 % of the patients misclassified.
These findings are valid for the conditions under which VV was assessed in our study, (i.e., VV assessed with the method of adjustments  in subacute hemispheric stroke patients seated, with the trunk and head maintained upright ) and for a fixed criterion (ICC ≥0.9), guaranteeing, in our opinion, a high reliability of the measure in clinical practice.
In our study, the minimum number of trials identified was the same for the two subgroups of patients with biased verticality perception (contralesional or ipsilesional), even if the underlying mechanisms of these biases are probably different . However, for similar SEMs for these two VV alterations, more trials (up to 10) are needed for patients presenting contralesional bias.
In patients with normal VV perception, VV orientation assessment required more than 10 trials to reach the ICC ≥0.9 cutoff. The lowest ICCs for this subgroup are certainly affected by the lower inter-subject variance (to −2.37° from 2.31°) as compared with other subgroups (to −14.72° from −2.61° and to 2.51° from 10.79° for contralesional and ipsilesional subgroups, respectively). Because ICC values are a ratio of the variance between subjects and the total variance, coefficients are systematically lower in homogeneous samples . This argument is supported by the correct SEM (1.13°) value with 10 trials for this subgroup. Considering the large size of this subgroup (n = 52) and that the inter-subject variance is limited by two boundaries, more than 10 trials may not contribute to significantly increasing the ICC.
In our study, participants adjusted the luminous rod without a time limit as is usual in clinical practice. This situation respects both the capacity and strategy of every patient. However, if healthy participants spend more time on single adjustments, the trial becomes more accurate and precise . Since the aim of this study was to generalize findings for clinical use in stroke rehabilitation, the procedure was similar to what we use in clinical practice.
We selected 10 trials as the maximum number of trials to assess VV perception in agreement with the literature on VV perception in stroke patients [2–4, 6, 7, 9] and to limit examination time due to fatigability in stroke patients. Fatigability was not quantified because the use of questionnaires is difficult with hemisphere stroke patients who frequently present aphasia and other cognitive disorders. However, the examiner ensured that patients felt able to complete the assessment and that their response did not become aberrant.