To investigate causes for reading comprehension decline with electronic devices, we prepared a repeated-measured design consisting of four conditions involving combinations of two mediums (smartphone and paper) and two sentences extracted from different texts (novels A and B; see Supplementary Text). One trial consisted of four sessions: resting state before reading, reading, resting state after reading, and reading test (Fig. 1a). Frontal brain activity (2 channels) were measured by functional Near-Infrared Spectroscopy (NIRS), and respiration activity (6 indexes) and metabolism patterns (2 indexes) were measured by respiratory Aeromonitor (see Methods).
For the reading score, repeated measured analysis of variance (RM-ANOVA) showed that reading medium affected reading scores (F1, 64 = 27.367, P < 0.0001, η2 = 0.297), whilst the main effects of sentence and the interaction were not. Post-hoc tests revealed that scores with the paper medium were higher than that with smartphone medium, in both novel sentences A and B (respectively P < 0.05, Fig. 1b). No main or interaction effects of medium or novel sentence were observed for duration of reading (Fig. 1c) or viewing distance between participants’ eyes and the device (Fig. 1d).
Out of 6 respiratory and 2 metabolic patterns that we measured, the tidal volume was decreased during reading compared to after and before reading sessions, and sighs were increased during reading with paper medium (Fig. 2a). RM-ANOVA showed that there were an effect of session but not medium and interaction for the tidal volume (Fig. 2b, session: F2,132 = 34.145, P < 0.0001, η2 = 0.341), inspiratory time (Fig. 2c, session: F2,132 = 8.639, P < 0.0001, η2 = 0.116), expiratory time (Fig. 2d, session: F2,132 = 33.092, P < 0.0001, η2 = 0.334), respiratory frequency (Fig. 2e, session: F2,132 = 29.018, P < 0.0001, η2 = 0.305), while there was an effect of session and interaction but not medium for the number of sighs (Fig. 2f, session: F2,132 = 20.530, P < 0.0001, η2 = 0.237; interaction: F2,132 = 9.169, P < 0.0001, η2 = 0.122). Post-hoc tests revealed that regardless of the medium, tidal volume, inspiratory time, and expiratory time were reduced during reading compared to resting states, whilst respiratory frequency was increased (all P < 0.05). Number of sighs was greater during reading with paper compared to smartphone (all P < 0.05). No changes in minute ventilation, O2 consumption, and end tidal CO2 were confirmed through all the session (Fig. 2g-i).
We found that activity recorded with NIRS increased during reading compared to before and after reading sessions, and was increased during reading from the smartphone medium, compared to the paper medium (Fig. 3a). The RM-ANOVA showed that no main effect of medium and interaction in channel 1 (left probe), while a main effect of session was significant (F2,132 = 19.612, P < 0.0001, η2 = 0.229). In the channel 2 (right probe), there were no main effect of medium and interaction, while a main effect of session was observed (F2,132 = 10.430, P < 0.0001, η2 = 0.136). Post-hoc tests revealed that the activity was increased during reading compared to resting states on both the left (P < 0.05, Fig. 3b) and right (P < 0.05, Fig. 3c) probes. Furthermore, the activity when reading from the smartphone was higher than when reading from paper on the left probe (all P < 0.05).
We conducted a path analysis to examine a mechanistic route for results of the reading tests. Six respiratory indexes, two metabolic indexes, two NIRS indexes, and reading score were set as the observed variables. In the most suitable model (Fig. 4, the goodness of fit index = 0.881), there are a covariate relationship between left NIRS channel activity and the number of sighs (P = 0.021), and a direct relationship between the right NIRS channel activity and its impact on reading score (P = 0.003).