ARTICLE   Open Access    

Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior

More Information
  • Previous research on the influence of media on prosocial behavior often focuses on the effects of watching TV/films, playing video games, and listening to music. Yet, less attention is paid to book reading, a traditional media use that continues to be prevalent, especially in adolescents' daily lives. Going beyond the specific content reading, this study explores the relationship between general book reading and the prosocial behavior of adolescents. Based on nationally representative data, Study 1 identified the positive impact of adolescents' book reading on their prosocial behavior. From a normative influence perspective, Study 2 validated the finding of Study 1 and investigated the underlying mechanism. Theoretically, these two studies extend the literature on the effects of media use on adolescents' prosocial behavior and highlight the role of normative influence in understanding this relationship. Practically, our findings are valuable references for practitioners in the book publishing industry and generate beneficial insights for adolescents' prosocial education.
  • Drought is a major abiotic stress affecting plant growth which becomes even more intensified as water availability for irrigation is limited with current climate changes[1]. Timely detection and identification of drought symptoms are critically important to develop efficient and water-saving irrigation programs and drought-tolerance turfgrasses. However, turfgrass assessments of stress damages have been mainly using the visual rating of turf quality which is subjective in nature and inclined to individual differences in light perception that drives inconsistency in estimating color, texture, and pattern of stress symptoms in grass species[24]. Remote sensing with appropriate imaging technology provides an objective, consistent, and rapid method of detecting and monitoring drought stress in large-scale turfgrass areas, which can be useful for developing precision irrigation programs and high-throughput phenotyping of drought-tolerance species and cultivars in breeding selection[5].

    Spectral reflectance and chlorophyll fluorescence imaging are emerging tools for rapid and non-destructive monitoring of drought effects in crops. These tools combine imaging and spectroscopy modalities to rigorously dissect the structural and physiological status of plants[6,7]. Spectral reflectance imaging captures reflected light (one out of three fates of light: reflect, absorb and transmit when striking leaf) at different wavelengths ranging from visible to near-infrared regions to characterize vegetation traits[8,9]. Within spectral reflectance imaging, multispectral imaging on one hand measures reflected light in three to ten different broad spectral bands in individual pixels[10,11]. Hyperspectral imaging on the other hand captures reflected light in narrow and more than 200 contiguous spectral bands. Some absorbed light by leaf is re-radiated back in the form of fluorescence and fluorescence imaging utilizes those lights in red and far-red regions to capture plant physiological status[12]. When drought progresses, plants start to develop various symptoms (physiological modifications) gradually over time[13]. Some of those symptoms include stomata closure, impediment in gas exchange, change in pigment composition and distribution which result in wilting and associated morphological alteration in leaf color (senescence), shape (leaf curling) and overall plant architecture. As different plant components or properties reflect light differently at different wavelengths and patterns of reflectance and fluorescence change along with plant stress and related symptoms development, spectral reflectance and fluorescence imaging provide accurate, reliable and detailed information for crop drought monitoring. Fluorescence imaging primarily based on fluorescing plant components or chlorophyll complex in photosynthetic antenna and reaction centers and therefore it mainly monitors stress development by tracking changes in overall photosynthetic performance or other metabolism that interfere with photosynthetic operation[9,14]. Multispectral imaging, hyperspectral imaging, or chlorophyll fluorescence has been used in different studies for plant responses to drought stress in various plant species[10,1517]. The comparative approach of multiple imaging technologies could help to find the efficient methods for the evaluation of plant responses and tolerance to drought[18].

    Vegetation indices derived from multispectral or hyperspectral imaging and fluorescence parameters typically are ratio or linear combinations of reflectance and fluorescence emissions from leaves or canopy of plants, respectively[19,20]. Canopy reflectance at different wavelengths and chlorophyll fluorescence varies with canopy color and density and changes with environmental conditions that affect plant growth, including drought stress[14,20,21]. These variations in reflectance and fluorescence are captured by vegetation indices, such as normalized difference vegetation index (NDVI) and fluorescence parameters including the ratio of variable fluorescence to maximum fluorescence (Fv/m) which are commonly used to evaluate environmental impact on plant growth. Other indices reflect physiological health of plants, such as photochemical reflectance index (PRI) has recently been reported to be useful for drought stress assessment in crops[19]. Previous research identified varying sensitivity of PRI and NDVI to detect water stress; for example, Sun et al.[22] found PRI to be a prominent indicator of drought stress whereas Kim et al.[20] discovered NDVI had greater correlation with drought stress development. There are also several conflicting findings on the responsiveness of fluorescence parameters to drought stress. Photochemical efficiency of PSII (Fv/Fm) was found to be greatly related to drought stress by Panigada et al.[23] but Jansen et al.[24] reported Fv/Fm to be relatively insensitive to drought progression. Lu & Zhang[25] identified that coefficient of photochemical quenching (qP) was insensitive to drought stress whereas Moustakas et al.[26] reported that (qP) being the most sensitive indicator of such stress conditions. There is a need for a comprehensive study that examines multiple vegetation indices (both hyperspectral and multispectral indices) and fluorescence parameters, and parallelly assess their sensitivities to reflect plant growth and physiological status during drought stress.

    The objectives of the current study were: (1) to perform comparative analysis of drought responses of vegetation and photosynthetic indices using multispectral, hyperspectral and chlorophyll fluorescence imaging for Kentucky bluegrass (Poa pratensis L.), a cool-season perennial grass species widely used as turfgrass; (2) identify major vegetation and photosynthetic indices from the imaging technologies and correlated to visual turf quality and leaf relative water content from the destructive measurement; and (3) determine the major vegetation and photosynthetic indices that are most responsive or sensitive to the progression of drought stress that may be useful to early detection and monitoring the level of drought stress causing growth and physiological damages in cool-season grass species.

    Sod strips of Kentucky bluegrass cultivar 'Dauntless' were collected from established field plots at the Rutgers Plant Science Research and Extension Farm, Adelphia, NJ, USA. Sods were planted in plastic pots of 18 cm diameter and 20 cm length filled with a mixture of soil (sandy loam, semi-active, mesic Typic Hapludult; pH 6.55; 260 kg·P·ha−1, 300 kg·K·ha−1) and sand in the ratio of 2/1 (v/v). Plants were established for 50-d in a greenhouse with 24/22 °C day/night average temperatures, 12-h average photoperiod and 750 μmol·m−2·s−1 average photosynthetically active radiation (PAR) with natural sunlight and supplemental lightings. Plants were well-watered, trimmed weekly to 100 mm and fertilized weekly with a 24–3.5–10 (N–P–K) fertilizer (Scotts Miracle-Gro) at the rate of 2.6 g·N·m−2 during the establishment period in the greenhouse. Once plants were well-established, they were moved to the controlled environmental growth chamber (GC72, Environmental Growth Chambers, Chagrin Falls, OH, USA). The growth chamber was controlled at 22/18 °C day/night temperature, 60% relative humidity, 12-h photoperiod and 650 μmol·m−2·s−1 PAR at the canopy level. Plants were allowed to acclimate for a week within the growth chamber conditions and then treatments were initiated.

    There were two different treatments: well-watered control and drought stress. For the well-watered control, plants were irrigated once every two days with sufficient water until drainage occurred from the pot bottom or when soil water content reached the field capacity. Drought stress was imposed by withholding irrigation from each pot throughout the experiment period. Each treatment had five replicates. The experimental treatments were arranged as a complete randomized design with plants of both treatments randomly placed and relocated in the chamber twice each week to minimize effects of potential microenvironment variations in the growth chamber.

    A time-domain reflectometry system (Model 6050 × 1; Soil Moisture Equipment, Santa Barbara, CA, USA) installed with 20 cm soil moisture probe was used to measure soil volumetric water content. Volumetric water content was measured every two days in each pot to track soil moisture dynamics in control and drought stress treatments. To assess plant responses at different soil moisture levels, turfgrass quality (TQ) and leaf relative water content (RWC) were evaluated. Turfgrass quality was visually rated on a scale of 1-9 depending upon canopy color, uniformity and density[27]. A rating of 1 indicates discolored and completely dead plants, 9 indicates lush green colored healthy plants and 6 indicates the minimum acceptable turfgrass quality. Leaf RWC was measured by soaking 0.2 g fresh leaves in distilled water overnight at 4 °C[28]. Turgid leaves after overnight soaking were oven dried at 70 °C to a constant dry weight. Leaf RWC was calculated as [(fresh weight – dry weight)/ (turgid weight – dry weight)] × 100.

    Control and drought stress pots were scanned using a close-range benchtop hyperspectral imaging system (Resonon Inc., Bozeman, MT, USA) containing Pika XC2 camera equipped with 23 mm lens. This camera took images in spectral range of 400–1,000 nm with much detailed spectral resolution of 1.9 nm in 447 different spectral channels. The camera provided 1600 spatial pixels and maximum frame rate of 165 frames per second. It had 23.1° field of view and 0.52 milli-radians instantaneous field of view. Resonon hyperspectral imaging systems are line-scan imagers (also referred to as push-broom imagers) that collect spectral data from each pixel on one line at a time. Multiple lines are imaged when an object or pot kept in scanning stage of linear stage assembly underneath the camera is moved by a stage motor. Those line images are assembled to form a complete image. The systems had regulated lights placed above the linear stage assembly to create optimal conditions for performing the scans. Lights were at the same level as the lens on a parallel plane. Distance between lens and the top of grass canopy was maintained at 0.4 m for capturing the best representation of drought progression. All scans were performed using spectronon pro (Resonon Inc., Bozeman, MT, USA) software connected to the camera using a USB cable. Before performing a scan, the lens was appropriately focused, dark current noise was removed and the system was calibrated for reflectance measurement using a white tile provided by the manufacturer. To ensure distortion-free hyperspectral datacube with a unit-aspect-ratio image, benchtop system's swatch settings were adjusted using pixel aspect ratio calibration sheet also provided by the manufacturer. Once the system was ready, controlled- and stressed-pots were scanned individually every two days throughout the experiment. As the lens was focused centrally, obtained images were of the central grass area and were processed using spectronon pro data analysis software. The entire grass image was selected using a selection tool and the spectrum was generated. From each spectrum, vegetation indices were calculated either using built-in plugins or by manually creating algorithms. The list of vegetation indices calculated using image analysis is mentioned in Table 1.

    Table 1.  List of vegetation indices calculated using hyperspectral and multispectral image analysis for drought stress monitoring in Kentucky bluegrass. Name and number in subscript following the letter R in each formula represent the reflectance at individual light and particular wavelength.
    Vegetation indexIndex abbreviation and formula
    Hyperspectral analysisMultispectral analysis
    Structure Independent Pigment IndexSIPI = (R800 – R445) / (R800 + R680)SIPIm = (RNIR840 – RBlue444) / (RNIR840 + RRed668)
    Simple Ratio IndexSRI = R800 / R675SRIm = RNIR840 / RRed668
    Plant Senescence Reflectance IndexPSRI = (R680 –R500) / R750PSRIm = (RRed668 – RBlue475) / RRededge740
    Photochemical Reflectance IndexPRI = (R570 – R531) / (R570 + R531)PRI = (RGreen560 – RGreen531) / (RGreen560 + RGreen531)
    Normalized Difference Vegetation IndexNDVI = (R800 – R680) / (R800 + R680)NDVIm = (RNIR840 – RRed668) / (RNIR840 + RRed668)
    Normalized Difference Red EdgeNDRE = (R750 – R705) / (R750 + R705)NDREm = (RRededge717 – RRed668) / (RRededge717 + RRed668)
     | Show Table
    DownLoad: CSV

    Micasense Rededge-MX dual camera system (AgEagle Sensor Systems Inc., Wichita, KS, USA) was used to collect multispectral images of controlled- and drought stressed-pots placed within a light box (1.2 m × 0.6 m × 0.6 m). The multispectral camera system had 1,280 × 960 resolution, 47.2° field of view and 5.4 mm focal length. The camera captured ten different spectral bands simultaneously on a command (Table 2). To allow the multispectral camera system, which was designed for aerial operation, to work in the light box settings, a downwelling light sensor (DLS) module provided by the manufacturer was installed to the camera system. Images were captured manually through WIFI connection from mobile devices or computer to the multispectral camera system. The sensor layout of the dual camera system, while causing negligible error in aerial condition, led to mismatching between spectral bands in a close distance, therefore, spectral bands needed to be overlapped during post-processing. The captured images of individual spectral bands were stored as separate .jpgf image files and then were used to calculate the relevant vegetation indices. Multispectral image analysis was executed using Python (Version 3.10) code by Rublee et al.[29]. Image analysis aligned ten spectral bands using Oriented FAST and Rotated BRIEF algorithm to achieve complete overlap between spectral band images. The reflectance correction panel provided by the manufacturer was used to account for the illumination condition in light box environment and the correction was reflected in pixel value adjustment for each band in python code; vegetation indices based on the aligned images were then calculated using the corresponding formula (Table 1). Images that included background noise were excluded from analysis.

    Table 2.  Spectral band details (center wavelength and band width) for Micasense Rededge-MX dual camera system.
    Band nameCentral wavelength (nm)Band width (nm)
    Blue44444428
    Blue47547532
    Green53153114
    Green56056027
    Red65065016
    Red66866814
    RE70570510
    RE71771712
    RE74074018
    NIR84084257
     | Show Table
    DownLoad: CSV

    Chlorophyll fluorescence images were taken using a pulse amplitude modulated fluorescence imaging system (FC 800-O/1010, Photon System Instruments, Drasov, Czech Republic). A high-speed charge-coupled device (CCD) camera was mounted on a robotic arm placed in the middle of LED light panels. The camera had 720 × 560 pixels spatial resolution, 50 frames per second frame rate and 12-bit depth. Four different LED light panels each of 20 cm × 20 cm size were equipped with 64 orange-red (617 nm) LEDs in three panels and 64 cool-white LEDs (6,500 k) in the rest of one panel. Before making measurements, plants were dark-adapted for 25 min in a dark room to open all PSII reaction centers. The distance between camera and the top of the grass canopy was maintained at 0.3 m while taking images to ensure optimum quality. Images were acquired following the Kautsky effect measured in a pulse amplitude modulated mode[30,31]. Briefly, dark-adapted plants were first exposed to non-actinic measuring light for 5 s to measure minimum fluorescence at the dark-adapted state (Fo). Plants were immediately exposed to 800 ms saturation pulse of 3,350 µmol·m−2·s−1 to measure maximum fluorescence after dark adaptation (Fm). They were kept under dark relaxation for 17 s and then exposed to actinic light 750 µmol·m−2·s−1 for 70 s. Plants were exposed to a series of saturating pulses at 8 s, 18 s, 28 s, 48 s and 68 s during their exposure to actinic light conditions and maximum fluorescence at different light levels and steady state were measured. They were kept under dark relaxation again for 100 s and irradiated with saturating pulses at 28 s, 58 s and 88 s during dark relaxation for measuring maximum fluorescence during the relaxation. Selected durations for each light and dark relaxation state were preset in default quenching-act2 protocol of the fluorescence imaging system. Fluorescence at different light levels and steady states were used to calculate several fluorescence parameters (Table 3).

    Table 3.  Chlorophyll fluorescence parameters calculated from pulse amplitude modulated fluorescence imaging system.
    Chlorophyll fluorescence parameterFormula
    Maximum photochemical efficiency of PSII (Fv / Fm)(Fm-Fo) / Fm
    Photochemical efficiency of open PSII centers
    (F'v / F'm)
    (F'm – F'o) / F'm
    Actual photochemical quantum yield of PSII centers Y(PSII)(F'm – Fs) / F'm
    Photochemical quenching coefficient (Puddle model; qP)(F'm – Fs) / (F'm – F'o)
    Photochemical quenching coefficient (Lake model; qL)qP × F'o / Fs
    Non-photochemical quenching coefficient (qN)(Fm-F'm) / Fm
    Non-photochemical quenching (NPQ)(Fm-F'm) / F'm
    Chlorophyll fluorescence decrease ratio (Rfd)(Fm-Fs) / Fs
     | Show Table
    DownLoad: CSV

    The two-way repeated measure analysis of variance was performed to determine treatment effects and t-test was performed to compare control and drought stress treatments at a given day of measurement. Correlation analysis using all individual observations (five replications for each control and drought stress treatments) was performed to determine the relationship among all measured traits, vegetation indices and fluorescence parameters. Partial least square regression (PLSR) models were developed in SAS JMP (version 13.2; SAS Institute, Cary, NC, USA) for comparing hyperspectral, multispectral and chlorophyll fluorescence imaging in their overall associations with physiological assessments of drought stress. Vegetation indices and fluorescence parameters from individual imaging technologies were predictor variables, and turfgrass quality and leaf relative water content were response variables. A leave one out cross validation approach was used to develop the best performing partial least square model for each imaging technology. A model was first established with all predictor variables and the variable with the lowest importance was removed from the dataset and the model was rebuilt with the remaining variables. The rebuilt model was re-validated using leave one out cross validation and assessed checking root mean PRESS and percent variation explained for cumulative Y values. From each loop of operation, one variable was removed, and a new model was developed. The whole process ended when the last variable was removed and thus no more models could be developed. Finally, a series of models was obtained, and they were compared to identify a model with the highest accuracy for individual imaging technologies. The best performing model from each imaging technology was used to estimate turfgrass quality and leaf relative water content.

    The initial soil water content prior to drought stress was maintained at the field capacity of 29% and remained at this level in the well-watered control treatment during the entire experimental period (20 d) (Fig. 1a). SWC in the drought treatment significantly decreased to below the well-watered treatment, beginning at 4 d, and declined to 5.8% by 20 d.

    Figure 1.  Drought stress affected turf quality, leaf relative water content and soil volumetric water content during 20 d of stress period in Kentucky bluegrass. * indicates significant difference between control and drought stress treatments (p ≤ 0.05) at each day of measurement. Presented values represent average of five data points.

    Leaf RWC was ≥ 93% in all plants prior to drought stress and declined to a significantly lower level than that of the control plants, beginning at 10 d of treatment when SWC declined to 16% (Fig. 1b). TQ began to decrease to a significantly lower level than the that of the well-watered plants at 10 d of drought stress at RWC of 87% and SWC of 16%, and further declined to the minimally acceptable level of 6.0 at 16 d of drought stress when RWC decreased to 66% and SWC dropped to 8% during drought stress (Fig. 1c).

    Most hyperspectral imaging indices, including SIPI (Fig. 2a), SRI (Fig. 2b), PRI (Fig. 2d), NDVI (Fig. 2e) and NDRE (Fig. 2f) exhibited a declining trend during 20-d drought stress while PSRI (Fig. 2C) showed increases during drought stress. The index value of drought-stressed plants became significantly lower than that of the well-watered plants, beginning at 14 d for SIPI and SRI, 16 d for PRI and PSRI, and 18 d for NDVI and NDRE. The multispectral SIPIm and SRIm did not differ between drought-stressed plants from the control plants until 18 d of treatment (Fig. 3a, b) while NDVIm, NDREm , PRIm , and PSRIm values were significantly lower than those of well-watered control plants at 16 d of drought stress (Fig. 3cf).

    Figure 2.  Vegetation indices generated by hyperspectral sensing and sensitivity of these indices in monitoring drought in Kentucky bluegrass exposed to 20 d of drought stress. * indicates significant difference between control and drought stress treatments (p ≤ 0.05) at each day of measurement. Presented values represent average of five data points.
    Figure 3.  Vegetation indices generated by multispectral image analysis and sensitivity of these indices in monitoring drought in Kentucky bluegrass exposed to 20 d of drought tress. * indicates significant difference between control and drought stress treatments (p ≤ 0.05) at each day of measurement. Presented values represent average of five data points.

    Chlorophyll fluorescence indices detected drought damages in leaf photosynthesis systems, as shown by declines in different indices during drought stress (Fig. 4). Drought-stressed plants exhibited significant lower chlorophyll fluorescence levels than that of the well-watered plants, beginning at 12 d for NPQ (Fig. 4a), 16 d for Fv/Fm (Fig. 4b), and 18 d for F'V/F'm (Fig. 4c), Y(PSII) (Fig. 4d), qP (Fig. 4e), and qL (Fig. 4f). Separation between drought-stressed and well-watered plants were also evident in index- or parameter- generated images (Fig. 5).

    Figure 4.  Chlorophyll fluorescence parameters measured by pulse amplitude modulated fluorescence imaging system and detection of drought by these parameters in Kentucky bluegrass exposed to 20 d of drought stress. * indicates significant difference between control and drought stress treatments (p ≤ 0.05) at each day of measurement. Presented values represent average of five data points. NPQ, Non-photochemical quenching; Fv /Fm, Maximum photochemical efficiency of PSII; F'v/F'm, Photochemical efficiency of open PSII centers; Y(PSII), Actual photochemical quantum yield of PSII centers; qP, Photochemical quenching coefficient (Puddle model); qL, Photochemical quenching coefficient (Lake model); qN, Non-photochemical quenching coefficient; Rfd, Chlorophyll fluorescence decrease ratio.
    Figure 5.  Maps generated by the three most drought sensitive indices and parameters [hyperspectral structure independent pigment index (SIPI), multispectral normalized difference vegetation index (NDVIm) and chlorophyll fluorescence NPQ]. These maps clearly separated control and drought stress after 18 d of treatment when majorities of indices and parameters detected drought stress.

    Leaf RWC and TQ had significant correlation with most of indices and parameters calculated using three different imaging sensors (hyperspectral, multispectral and chlorophyll fluorescence) (Table 4). Among the indices, RWC had the strongest correlations with chlorophyll fluorescence NPQ (r = 0.88) and qL (r = 0.89), hyperspectral PRI (r = 0.94), and multispectral PSRIm (−0.92). TQ was most correlated to chlorophyll fluorescence NPQ (r = 0.89), hyperspectral PSRI (r = −0.90), and multispectral PSRIm (r = −0.85).

    Table 4.  Correlations among several physiological traits, vegetation indices and chlorophyll fluorescence parameters.
    RWCTQFV/FmF'v/F'mY(PSII)NPQqNqPqLRfdSIPISRIPSRIPRINDVINDREWBISIPImPSRImPRImNDVImNDREm
    RWC1.00
    TQ0.95*1.00
    FV/Fm0.87*0.85*1.00
    F'v/F'm0.81*0.77*0.95*1.00
    Y(PSII)0.85*0.74*0.80*0.74*1.00
    NPQ0.88*0.89*0.95*0.84*0.75*1.00
    qN0.84*0.83*0.96*0.84*0.77*0.96*1.00
    qP0.82*0.70*0.73*0.66*0.99*0.69*0.72*1.00
    qL0.89*0.81*0.90*0.86*0.97*0.83*0.86*0.95*1.00
    Rfd0.84*0.82*0.89*0.83*0.77*0.92*0.86*0.72*0.83*1.00
    SIPI0.84*0.71*0.63*0.58*0.51*0.57*0.69*0.48*0.60*0.46*1.00
    SRI0.57*0.62*0.44*0.45*0.330.41*0.45*0.300.400.330.83*1.00
    PSRI−0.83*−0.90*−0.90*−0.86*−0.76*−0.83*−0.87*−0.71*−0.86*−0.76*−0.75*−0.57*1.00
    PRI0.94*0.82*0.80*0.76*0.71*0.79*0.71*0.66*0.77*0.78*0.260.17−0.78*1.00
    NDVI0.53*0.65*0.41*0.43*0.41*0.42*0.400.380.43*0.42*0.50*0.42*−0.54*0.311.00
    NDRE0.64*0.73*0.64*0.63*0.45*0.54*0.64*0.400.56*0.44*0.92*0.85*−0.75*0.330.50*1.00
    SIPIm0.52*0.50*0.56*0.58*0.47*0.52*0.49*0.43*0.52*0.51*0.330.28−0.58*0.61*0.270.39−0.281.00
    PSRIm−0.92*−0.85*−0.85*−0.85*−0.83*−0.80*−0.77*−0.79*−0.88*−0.77*−0.40−0.230.77*−0.82*−0.41−0.400.32−0.52*1.00
    PRIm0.20−0.030.06−0.010.280.140.110.310.200.180.050.100.01−0.040.000.090.060.09−0.041.00
    NDVIm0.75*0.74*0.77*0.78*0.67*0.72*0.68*0.62*0.73*0.70*0.43*0.33−0.76*0.81*0.370.47*−0.350.93*−0.76*−0.051.00
    NDREm0.90*0.89*0.89*0.89*0.81*0.83*0.81*0.76*0.88*0.81*0.52*0.41*−0.87*0.87*0.45*0.53*−0.320.62*−0.87*−0.040.85*1.00
    Details for individual abbreviations of vegetation indices and fluorescence parameters used in this table were previously mentioned in Tables 1 & 3. Some other abbreviations are: RWC, leaf relative water content; and TQ, turfgrass quality. Values followed by * indicate significant correlation at p ≤ 0.05. Correlation analysis was performed using all individual data points (five replications for each control and drought stress treatments).
     | Show Table
    DownLoad: CSV

    Correlations among different vegetation indices and parameters were also significant in many cases. Hyperspectral indices such as PSRI and PRI were significantly correlated with all multispectral indices except PRIm. Multispectral NDVIm and NDREm were significantly correlated with all hyperspectral indices. When hyperspectral and multispectral indices were correlated with chlorophyll fluorescence parameters, majorities of these indices significantly associated with fluorescence parameters with exceptions of multispectral PRIm which had weak and positive (r ranges 0.06 to 0.31) associations with fluorescence parameters.

    Partial least square regression models were developed by integrating all indices from individual imaging technologies which identified the most reliable imaging systems to detect and monitor plant responses to drought stress. The PLSR model developed using hyperspectral imaging indices had improved predictability (root mean PRESS ≤ 0.38 and percent variation explained ≥ 87) compared to such models developed using other imaging systems and associated indices (Table 5). Comparing multispectral imaging with chlorophyll fluorescence imaging, multispectral imaging had slightly better predictability [root mean PRESS = 0.40 (RWC) and 0.44 (TQ) and percent variation explained = 86 (RWC) and 83 (TQ)] considering similar number of predictor variables used for estimating TQ and RWC in all imaging systems.

    Table 5.  Summary of partial least square model showing predictability of individual models using specific numbers of predictor variables (identified by leave one out cross validation) generated by different sensing technologies. Details of individual abbreviations are mentioned in previous tables. Partial least square was performed using all individual data points (five replications for each control and drought stress treatments).
    Sensing technology used for predictionPredicted
    variable
    No. of predictors usedPredictor variablesRoot mean
    PRESS
    Percent variation explained
    for cumulative Y
    Cumulative Q2
    HyperspectralTQ4PRI, PSRI, NDRE, SIPI0.36870.99
    RWC4PRI, PSRI, NDRE, SIPI0.38890.99
    MultispectralTQ3PSRIm, NDVIm, NDREm0.44850.97
    RWC3PSRIm, NDVIm, NDREm0.40860.97
    Chlorophyll fluorescenceTQ4Fv/Fm, NPQ, qN, qL0.46830.95
    RWC3Fv/Fm, NPQ, qL0.59840.93
     | Show Table
    DownLoad: CSV

    The integrated indices from each of the three imaging systems were highly correlated to TQ, with R2 of 0.90, 0.85, and 0.83 for hyperspectral imaging, multispectral imaging, and chlorophyll fluorescence, respectively (Fig. 6). For RWC, the correlation R2 was 0.88, 0.84, and 0.80, respectively with hyperspectral imaging, multispectral imaging, and chlorophyll fluorescence. The hyperspectral imaging was better be able to predict TQ and RWC compared to other imaging systems (Fig. 6).

    Figure 6.  Comparison of predicted turfgrass quality (TQ) and leaf relative water content (RWC) versus their measured values using partial least square regression model. Turfgrass quality and relative water contents were predicted using various indices generated by hyperspectral, multispectral and chlorophyll fluorescence sensing technologies. The dashed line represents the I:I line. Regression analysis was performed using all individual data points (five replications for each control and drought stress treatments).

    Leaf RWC and TQ are the two most widely used parameters or traits to evaluate turfgrass responses to drought stress[28,32,33]. In this study, RWC detected water deficit in leaves at 10 d of drought stress when SWC declined to 16% and TQ declined to below the minimal acceptable level of 6.0 at 16 d of drought stress when RWC decreased to 66% and SWC dropped to 8% during drought stress. These results suggested that RWC was a sensitive trait to detect water stress in plants, which is in agreement with previous research[34,35]. However, leaf RWC is a destructive measurement and TQ is a subjective estimate. Nondestructive and quantitative detection of stress symptoms in plants through assessing changes in phenotypic and physiological responses of plants to drought stress is critical for developing water-saving irrigation programs and breeding selection traits to increase water use efficiency and improve plant tolerance to drought stress. In this study, some of the phenotypic traits assessed by hyperspectral and multispectral imaging analysis and photosynthetic parameters measured by chlorophyll fluorescence were highly correlated to leaf RWC or visual TQ, as discussed in detail below, which could be used as non-destructive indicators or predictors for the level of drought stress in Kentucky bluegrass and other cool-season turfgrass species.

    The strong correlation of integrated indices from each of the three imaging systems with TQ (R2 of 0.90, 0.85, and 0.83, respectively) and RWC (R2 of 0.88, 0.84, and 0.80, respectively) for hyperspectral imaging, multispectral imaging, and chlorophyll fluorescence suggested that all three non-destructive imaging systems could be used as a non-destructive technique to detect and monitor water stress in Kentucky bluegrass. However, the hyperspectral imaging indices had higher predictability to RWC and visual TQ compared to the indices from multispectral imaging and chlorophyll fluorescence based on the PLSR models. Hyperspectral imaging used in this study captured images in 447 different spectral bands and gathered much more details about individual components of entire vegetation as each component has its own spectral signature. Multispectral imaging captures images with ten spectral bands and chlorophyll fluorescence imaging used only emitted red and far-red lights to snap images. Nevertheless, our results suggested that the PLSR models by integrating all indices from each individual imaging technologies identified the most reliable imaging systems to detect and monitor plant responses to drought stress in this study.

    The indices derived from the three imaging systems varied in their correlation to RWC or TQ in Kentucky bluegrass in this study. Among the indices, RWC had the strongest correlations with chlorophyll fluorescence NPQ (r = 0.88) and qL (r = 0.89), hyperspectral PRI (r = 0.94), and multispectral PSRIm (r = −0.92). TQ was most correlated to chlorophyll fluorescence NPQ (r = 0.89), hyperspectral PSRI (r = −0.90), and multispectral PSRIm (r = −0.85). The indices also varied in their sensitivity to drought stress for Kentucky bluegrass, and therefore they detected drought stress in plants at different times of treatment. The hyperspectral SIPI and SRI were the most responsive to drought stress with significant decline at 14 d followed by PRI and PSRI at 16 d while NDVI and NDRE were slowest showing decline (18 d) in response to drought. Multispectral indices exhibited decline later during drought at 16 d of drought stress for NDVIm, NDREm , PRIm , and PSRIm and 18 d for SIPIm and SRIm. Indices SIPI and SRI are related to leaf carotenoid composition and vegetation density and high spectral resolution of hyperspectral system was able to capture subtle changes in pigment concentration and canopy (slight leaf shrinking and rolling) at early phase of drought progression[36,37]. Index PSRI is indicative of the ratio of bulk carotenoids including α- and β-carotenes to chlorophylls and PRI is sensitive to xanthophyll cycle particularly de-epoxidation of zeaxanthin that releases excess energy as heat in order to photoprotection[3840]. Activation of photoprotective mechanisms including xanthophyl cycle require a certain level of stress severity depending on type of abiotic stress and plant species[41]. The PSRI calculated using both hyperspectral and multispectral imaging systems exhibited similar trends, and PSRI and PRI from either imaging system detected drought stress after 16 days of treatment applications. In agreement with our results, Das & Seshasai[42] found that PSRI showed similar trends when its value > −0.2 regardless of whether measured using hyperspectral or multispectral imaging. Both PSRI and PRI were also highly correlated to leaf RWC or TQ in Kentucky bluegrass exposed to drought stress in this study, suggesting that these two indices could be useful parameters to detect and monitor plant responses to drought stress.

    Vegetation index of NDVI has been the most widely used vegetation index in several crops such as wheat (Triticum aestivum L.)[43], cool- and warm-season turfgrass species including perennial ryegrass (Lolium perenne L.), tall fescue (Festuca arundinacea Schreb.), seashore paspalum (Paspalum vaginatum Sw.) and hybrid bermudagrass [Cynodon dactylon (L.) Pers. × C. transvaalensis Burtt-Davy][2, 44, 45]. For example, Bhandari et al.[43] and Badzmierowski et al.[14] found NDVI was correlated to overall turfgrass quality and chlorophyll content under nitrogen and drought stresses in tall fescue and citrus (Citrus spp.) plants. In this study, NDVI and NDRE were also correlated to leaf RWC and TQ, both NDVI and NDRE calculated from hyperspectral or multispectral imaging were least responsive to drought stress or detected drought stress later than other indices. Hong et al.[46] reported that NDVI being a better indicator than NDRE for early drought stress detection in turfgrasses when these indices were measured by handheld multispectral sensor. Eitel et al.[47] utilized broadband satellite images to estimate NDVI and NDRE and identified NDRE being a better option for early detection of stress condition in woodland area. Either NDVI or NDRE could be used as indices for vegetation density, but not sensitive indicators for plant responses to drought stress or for detecting drought damages in plants.

    Chlorophyll fluorescence reflects the integrity and functionality of photosystems in the light reactions of photosynthesis and serves as a good indicator for photochemical activity and efficiency[48]. The Y(PSII) is an effective quantum yield of photochemical energy conversion and estimates the actual proportion of absorbed light that is used for electron transport[49]. The ratio of F'v/F'm is maximum proportion of absorbed light that can be used for electron transport when all possible PSII reaction centers are open under light adapted state. Parameters qP and qL estimate the fraction of open PSII centers based on 'puddle' and 'lake or connected unit' models of photosynthetic antenna complex, respectively[50]. Rfd is an indicator for photosynthetic quantum conversion associated with functionality of the photosynthetic core unit. Overall, these parameters revolve around the operation status and functioning of PSII reaction centers or the core unit that possesses chlorophyll a-P680 in a matrix of proteins[51]. Parameter NPQ indicates non-photochemical quenching of fluorescence via heat dissipation involving xanthophyll cycle and state transition of photosystems[52]. This parameter is mostly associated with xanthophylls and other pigments in light harvesting antenna complex of photosystems but not with the PSII core unit[53]. Li et al.[9] reported that chlorophyll fluorescence imaging parameters including F'V/F'm have a limitation of late drought detection in plants. Shin et al.[54] reported F'V/F'm, Y(PSII), qP, and qL detected stress effects under severe drought when leaves were completely wilted and fresh weights declined in lettuce (Lactuca sativa L.) seedings. In this study, NPQ and Fv/Fm exhibited significant decline earlier (12−16 d of stress treatment) when drought was in mild to moderate level (> 60% leaf water content) compared to other chlorophyll fluorescence indices. The NPQ was strongly correlated to leaf RWC (r = 0.88) and TQ (r = 0.89) for Kentucky bluegrass exposed to drought stress. These results suggested that NPQ is a sensitive indicator of photosynthetic responses to drought stress and could be a useful parameter for evaluating plant tolerance to drought stress and monitoring drought responses.

    The comparative analysis of phenotypic and photosynthetic responses to drought stress using three imaging technologies (hyperspectral, multispectral and chlorophyll fluorescence) using the partial least square modeling demonstrated that the integrated vegetation indices from hyperspectral imaging had higher predictability for detecting turfgrass responses to drought stress relative to those from multispectral imaging and chlorophyll fluorescence. Among individual indices, SIPI and SRI from hyperspectral imaging were able to detect drought stress sooner than others while PSRI and PRI from both hyperspectral and multispectral imaging were also highly correlated to leaf RWC or TQ responses to drought stress, suggesting these indices could be useful parameters to detect and monitor drought stress in cool-season turfgrass. While NDVI or NDRE from both hyperspectral and multispectral imaging could be used as indices for vegetation density, but not sensitive indicators for plant responses to drought stress. Among chlorophyll fluorescence indices, NPQ and Fv/Fm were more closely correlated to RWC or TQ while NPQ was most responsive to drought stress, and therefore NPQ could be a useful indicator for detecting and monitoring cool-season turfgrass response to drought stress. The sensitivity and effectiveness of these indices associated with drought responses in this study could be further testified in other cool-season and warm-season turfgrass species under field conditions. As each imaging technology used in this experiment comes with bulky accessories such as LED panels, mounting tower and support system, capturing images within limited space of controlled environmental chambers are challenging. Future research should be in developing multimodal imaging integrating major features of all three technologies and reducing size and space requirement that would deliver improved decision support for drought monitoring and irrigation management in turfgrasses. Development of advanced algorithms that could incorporate broader spectral details or band reflectance for calculating novel vegetation indices are warranted.

    The research presented in this paper was funded by the United State Department of Agriculture - National Institute of Food and Agriculture (2021-51181-35855).

  • The authors declare that they have no conflict of interest. Bingru Huang is the Editorial Board member of Journal Grass Research who was blinded from reviewing or making decisions on the manuscript. The article was subject to the journal's standard procedures, with peer-review handled independently of this Editorial Board member and her research groups.

  • [1]

    National Press and Publication Administration. 2022. Key Findings of the 19th National Reading Survey (In Chinese). www.nppa.gov.cn/nppa/contents/280/103913.shtml

    [2]

    Wang Z, Zhang X, Zhang G. 2021. Creating a ubiquitous reading environment for children in undeveloped rural areas: An action research project. Library & Information Science Research 43:101118

    doi: 10.1016/j.lisr.2021.101118

    CrossRef   Google Scholar

    [3]

    Wasik BA, Hindman AH, Snell EK. 2016. Book reading and vocabulary development: A systematic review. Early Childhood Research Quarterly 37:39−57

    doi: 10.1016/j.ecresq.2016.04.003

    CrossRef   Google Scholar

    [4]

    Bano J, Jabeen Z, Qutoshi SB. 2018. Perceptions of teachers about the role of parents in developing reading habits of children to improve their academic performance in schools. Journal of Education and Educational Development 5(1):42−59

    doi: 10.22555/joeed.v5i1.1445

    CrossRef   Google Scholar

    [5]

    Karadeniz A, Can R. 2015. A research on book reading habits and media literacy of students at the faculty of education. Procedia - Social and Behavioral Sciences 174:4058−67

    doi: 10.1016/j.sbspro.2015.01.1155

    CrossRef   Google Scholar

    [6]

    Eisenberg N, Fabes RA, Spinrad TL. 2006. Prosocial Development. In Handbook of child psychology: Social, emotional, and personality development , ed. Damon W, Lerner RM, Eisenberg N. Vol. 3, 6th edition. Hoboken, New Jersey: John Wiley & Sons, Inc. pp. 646-718.

    [7]

    Wentzel KR, Filisetti L, Looney L. 2007. Adolescent prosocial behavior: The role of self-processes and contextual cues. Child Development 78:895−910

    doi: 10.1111/j.1467-8624.2007.01039.x

    CrossRef   Google Scholar

    [8]

    Closson LM. 2009. Aggressive and prosocial behaviors within early adolescent friendship cliques: What's status got to do with it? Merrill-Palmer Quarterly 55:406−35

    doi: 10.1353/mpq.0.0035

    CrossRef   Google Scholar

    [9]

    Son D, Padilla-Walker LM. 2020. Happy helpers: A multidimensional and mixed-method approach to prosocial behavior and its effects on friendship quality, mental health, and well-being during adolescence. Journal of Happiness Studies 21:1705−23

    doi: 10.1007/s10902-019-00154-2

    CrossRef   Google Scholar

    [10]

    Gerbino M, Zuffianò A, Eisenberg N, Castellani V, Luengo Kanacri BP, et al. 2018. Adolescents' prosocial behavior predicts good grades beyond intelligence and personality traits. Journal of Personality 86:247−60

    doi: 10.1111/jopy.12309

    CrossRef   Google Scholar

    [11]

    Coyne SM, Padilla-Walker LM, Holmgren HG, Davis EJ, Collier KM, et al. 2018. A meta-analysis of prosocial media on prosocial behavior, aggression, and empathic concern: A multidimensional approach. Developmental Psychology 54:331−47

    doi: 10.1037/dev0000412

    CrossRef   Google Scholar

    [12]

    De Leeuw RNH, Kleemans M, Rozendaal E, Anschütz DJ, Buijzen M. 2015. The impact of prosocial television news on children's prosocial behavior: An experimental study in the Netherlands. Journal of Children and Media 9:419−34

    doi: 10.1080/17482798.2015.1089297

    CrossRef   Google Scholar

    [13]

    Greitemeyer T. 2009. Effects of songs with prosocial lyrics on prosocial thoughts, affect, and behavior. Journal of Experimental Social Psychology 45:186−90

    doi: 10.1016/j.jesp.2008.08.003

    CrossRef   Google Scholar

    [14]

    Ruth N. 2017. "Heal the World": A field experiment on the effects of music with prosocial lyrics on prosocial behavior. Psychology of Music 45:298−304

    doi: 10.1177/0305735616652226

    CrossRef   Google Scholar

    [15]

    Greitemeyer T, Mügge DO. 2014. Video games do affect social outcomes: A meta-analytic review of the effects of violent and prosocial video game play. Personality and Social Psychology Bulletin 40:578−89

    doi: 10.1177/0146167213520459

    CrossRef   Google Scholar

    [16]

    Prot S, Gentile DA, Anderson CA, Suzuki K, Swing E, et al. 2014. Long-Term Relations Among Prosocial-Media Use, Empathy, and Prosocial Behavior. Psychological Science 25:358−68

    doi: 10.1177/0956797613503854

    CrossRef   Google Scholar

    [17]

    Jian YC. 2022. Reading in print versus digital media uses different cognitive strategies: evidence from eye movements during science-text reading. Reading and Writing 35:1549−68

    doi: 10.1007/s11145-021-10246-2

    CrossRef   Google Scholar

    [18]

    Zhou S, Kiesow D, Guo L. 2023. The Values of Print: Affordances and Sensemaking for Newspaper Consumers. Journalism Practice 17:336−53

    doi: 10.1080/17512786.2021.1910984

    CrossRef   Google Scholar

    [19]

    van der Weel A, Mangen A. 2022. Textual reading in digitised classrooms: Reflections on reading beyond the internet. International Journal of Educational Research 115:102036

    doi: 10.1016/j.ijer.2022.102036

    CrossRef   Google Scholar

    [20]

    Li W, Zhou L, Ai P, Kim GR. 2022. Reading prosocial content in books and adolescents' prosocial behavior: A moderated mediation model with evidence from China. Frontiers in Psychology 13:973481

    doi: 10.3389/fpsyg.2022.973481

    CrossRef   Google Scholar

    [21]

    Mar RA, Oatley K, Hirsh J, dela Paz J, Peterson JB. 2006. Bookworms versus nerds: Exposure to fiction versus non-fiction, divergent associations with social ability, and the simulation of fictional social worlds. Journal of Research in Personality 40:694−712

    doi: 10.1016/j.jrp.2005.08.002

    CrossRef   Google Scholar

    [22]

    Mar RA, Oatley K, Peterson JB. 2009. Exploring the link between reading fiction and empathy: Ruling out individual differences and examining outcomes. Communications 34(4):407−28

    doi: 10.1515/comm.2009.025

    CrossRef   Google Scholar

    [23]

    Kiesow D, Zhou S, Guo L. 2023. Affordances for Sense-Making: Exploring Their Availability for Users of Online News Sites. Digital Journalism 11:962−81

    doi: 10.1080/21670811.2021.1989316

    CrossRef   Google Scholar

    [24]

    Chen Y, Yan F, He G, Yan W. 2020. The werther effect revisited: Do suicides in books predict actual suicides? Poetics 81:101441

    doi: 10.1016/j.poetic.2020.101441

    CrossRef   Google Scholar

    [25]

    Bavishi A, Slade MD, Levy BR. 2016. A chapter a day: Association of book reading with longevity. Social Science & Medicine 164:44−48

    doi: 10.1016/j.socscimed.2016.07.014

    CrossRef   Google Scholar

    [26]

    Sikora J, Evans MDR, Kelley J. 2019. Scholarly culture: How books in adolescence enhance adult literacy, numeracy and technology skills in 31 societies. Social Science Research 77:1−15

    doi: 10.1016/j.ssresearch.2018.10.003

    CrossRef   Google Scholar

    [27]

    Lehalle H. 2007. Cognitive development in adolescence: Thinking freed from concrete constraints. In Handbook of adolescent development, eds. Jackson S, Goossens L. 1st Edition. London: Psychology Press. pp. 71−89. https://doi.org/10.4324/9780203969861-5

    [28]

    Xiao D, Yu W. 2010. Global History of Publishing (Zhongwai Chuban Shi). Beijing, China: China Renmin University Press. 215 pp.

    [29]

    Li W, Ai P, Zhou D. 2020. More reading, more pro-social? Evidence from 2012 Chinese general social survey China Publishing Journal 2020(16):8−14

    doi: 10.3969/j.issn.1002-4166.2020.16.002

    CrossRef   Google Scholar

    [30]

    Schwartz SH. 1977. Normative influences on altruism. Advances in Experimental Social Psychology 10:221−79

    doi: 10.1016/S0065-2601(08)60358-5

    CrossRef   Google Scholar

    [31]

    Yang Y, Zhang M, Kou Y. 2016. The Revalidation and Development of the Prosocial Behavior Scale for Adolescent. Chinese Social Psychological Review 10:135−50

    Google Scholar

    [32]

    Ellison NB, Steinfield C, Lampe C. 2007. The benefits of facebook "Friends:" social capital and college students' use of online social network sites. Journal of Computer-Mediated Communication 12:1143−68

    doi: 10.1111/j.1083-6101.2007.00367.x

    CrossRef   Google Scholar

    [33]

    Ajzen I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50:179−211

    doi: 10.1016/0749-5978(91)90020-T

    CrossRef   Google Scholar

    [34]

    Smith JR, McSweeney A. 2007. Charitable giving: the effectiveness of a revised theory of planned behaviour model in predicting donating intentions and behaviour. Journal of Community & Applied Social Psychology 17:363−86

    doi: 10.1002/casp.906

    CrossRef   Google Scholar

    [35]

    Carlo G, Randall BA. 2002. The development of a measure of prosocial behaviors for late adolescents. Journal of Youth and Adolescence 31:31−44

    doi: 10.1023/A:1014033032440

    CrossRef   Google Scholar

    [36]

    Hayes AF. 2022. Introduction to mediation, moderation, and conditional process analysis : a regression-based approach. 3rd Edition. London, New York: The Guilford Press

    [37]

    Preacher KJ, Hayes AF. 2004. SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, Instruments & Computers 36:717−31

    doi: 10.3758/BF03206553

    CrossRef   Google Scholar

    [38]

    Hayes AF, Rockwood NJ. 2017. Regression-based statistical mediation and moderation analysis in clinical research: Observations, recommendations, and implementation. Behaviour Research and Therapy 98:39−57

    doi: 10.1016/j.brat.2016.11.001

    CrossRef   Google Scholar

    [39]

    Larsen NE, Lee K, Ganea PA. 2018. Do storybooks with anthropomorphized animal characters promote prosocial behaviors in young children? Developmental Science 21:e12590

    doi: 10.1111/desc.12590

    CrossRef   Google Scholar

    [40]

    Greitemeyer T, Osswald S. 2010. Effects of prosocial video games on prosocial behavior. Journal of Personality and Social Psychology 98:211−21

    doi: 10.1037/a0016997

    CrossRef   Google Scholar

    [41]

    Li H, Zhang Q. 2023. Effects of prosocial video games on prosocial thoughts and prosocial behaviors. Social Science Computer Review 41(3):1063−80

    doi: 10.1177/08944393211069599

    CrossRef   Google Scholar

  • Cite this article

    Ai P, Li W. 2023. Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior. Publishing Research 2:5 doi: 10.48130/PR-2023-0005
    Ai P, Li W. 2023. Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior. Publishing Research 2:5 doi: 10.48130/PR-2023-0005

Figures(1)  /  Tables(4)

Article Metrics

Article views(3345) PDF downloads(570)

Other Articles By Authors

ARTICLE   Open Access    

Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior

Publishing Research  2 Article number: 5  (2023)  |  Cite this article

Abstract: Previous research on the influence of media on prosocial behavior often focuses on the effects of watching TV/films, playing video games, and listening to music. Yet, less attention is paid to book reading, a traditional media use that continues to be prevalent, especially in adolescents' daily lives. Going beyond the specific content reading, this study explores the relationship between general book reading and the prosocial behavior of adolescents. Based on nationally representative data, Study 1 identified the positive impact of adolescents' book reading on their prosocial behavior. From a normative influence perspective, Study 2 validated the finding of Study 1 and investigated the underlying mechanism. Theoretically, these two studies extend the literature on the effects of media use on adolescents' prosocial behavior and highlight the role of normative influence in understanding this relationship. Practically, our findings are valuable references for practitioners in the book publishing industry and generate beneficial insights for adolescents' prosocial education.

    • Despite the rapid change in media use, book reading, a relatively traditional media use, is still prevalent and prominent in our modern lives, especially for adolescents. According to a nationwide survey in China in 2021, the book reading rate among adolescents reached over 90% and is the highest among all age groups[1]. Reading books has a life-long impact on people[2]. Therefore, such a high book reading rate among adolescents suggests that research on the influence of book reading on this group is warranted. Studies have shown that book reading is predominantly beneficial for adolescents and young adults at the intrapersonal level, such as developing their vocabulary[3], improving their academic performance[4], and increasing their media literacy[5]. However, the potential influence of book reading on the interpersonal and social aspects of adolescents has not been fully explored.

      One of the most prominent concerns in the interpersonal and social development of adolescents is prosocial behavior. Prosocial behavior is an important category of human behavior operated at interpersonal and societal levels, which in general refers to 'behaviors that benefit other people, such as helping, sharing, donating and volunteering[6]. For adolescents, performing prosocial behavior is a hallmark of their social competence[7] and benefits them in various aspects, such as improving their friendship quality[8], well-being[9] and academic performance[10]. Therefore, in past decades, scholars from various fields have explored the different factors stimulating prosocial behavior to better nurture such behavior, among which media use is one of the essential factors[11].

      Although research has tested prosocial outcomes induced by the usage of several types of media, including television[12], music[13,14], and video games,[15,16] much less attention has been paid to books, a traditional media yet still prevalent and prominent in our modern lives. In today's digital era, heightened worries about excessive screen time and its potential impact on health and well-being have sparked a keen interest in pursuits such as book reading, which offer a refreshing alternative[17,18]. Furthermore, the significance of reading books holds a pivotal role in the realm of education[19]. Among the limited published literature on book reading and prosocial outcomes, most of them basically followed the traditional research approach adopted by media study scholars, which focuses on readers' exposure to specific prosocial content[20] or specific genre[21,22]. These studies lay down a good foundation for understanding how book reading is associated with prosocial behavior. Nevertheless, specifying the content of the books lacks ecological and external validity because the books that people read in daily life contain various genres of books.

      In the current study, we argue that more general book reading that does not specify the type of content could also predict adolescents' prosocial behavior from the perspective of media affordance. The affordance of print media (e.g., books and newspapers) reduce distractions and enable its readers to take more time to browse and think, facilitating the apprehension process[18,23]. Furthermore, unlike other text in other media forms like social media posts, newspapers, or magazines, the content in books is much longer in length and deeper in content[24,25]. Consequently, reading books involves a higher cognitive load and can better enhance cognitive development and literacy than other media types[25,26]. Adolescents are in the stage of rapid cognitive development[27]. As such, the effect of reading books on cognitive abilities could be more prominent in this group.

      At the same time, due to the strict gatekeeping in terms of book publication in almost all countries and areas[28], books generally contain more content concerning positive social norms about prosocial behavior compared to other media. This is especially true in China since the Chinese government particularly emphasizes the educational duty of books, aiming at improving people's morality and prosociality via book reading[29]. Moreover, given the augmentation function of cognitive abilities through reading books, readers thereby could be more capable of learning and understanding social norms conveyed in books. Also, social norms can be internalized into personal norms, which is an essential antecedent of prosocial behavior[30]. Taken together, we argue that general book reading can improve adolescents' cognitive abilities that drive the effect of book reading on prosocial behavior via the mediating roles of social norms and personal norms.

      Based on these arguments, we conducted two studies with different emphases. Study 1 seeks to identify the association between adolescents' book reading and their prosocial behavior with nationally representative data. Study 2 then collects survey data with more comprehensive measures to further validate the finding of Study 1 and explores the underlying mechanisms. Specifically, we develop a model through the normative influence's perspective, examining social norms and personal norms as two serial mediators in the relationship between adolescents' book reading and prosocial behavior. Theoretically, these two studies go beyond the focus on exposure to specific reading content or genres to explore the impact of general book reading on individuals' prosocial behavior, which enriches the current literature on the prosocial outcomes of media use. Practically, our findings are valuable references for practitioners in the book publishing industry and generate beneficial insights for adolescents' prosocial education.

    • To explore whether general book reading has an impact on prosocial behavior among adolescents, we resorted to China Education Panel Survey (CEPS), a nationally representative panel data in China. Because the data is publicly available, the ethical review was exempted by the Institutional Review Board of the corresponding author's institute. In the 2013−2014 academic year, CEPS surveyed 19,487 students (10,279 seventh-grade students and 9,208 ninth-grade students) from 112 schools in 28 county-level areas in China. This study used the data of seventh-grade students, among which 9,449 completed the follow-up survey during 2014−2015 academic year. CEPS not only included the survey of students themselves, but their important others such as parents. Among all the participants, 47.02% were females, and 52.80% were males. The age range of the participants was 11 to 18 (M = 12.967, SD = 0.894).

    • The dependent variable is prosocial behavior. The measurement of this variable was from the follow-up survey by asking the students the frequency of three prosocial behaviors during the past year. The three prosocial behaviors were (1) helping the elders, (2) being kind and friendly to others, and (3) following rules and not cutting in lines (1 = never, 5 = always; M = 3.789, SD = 0.771, Cronbach's α = 0.680). Our selection of the three behaviors was primarily guided by the prosocial scale for Chinese adolescents formulated by Yang et al. taking into consideration the cultural nuances of prosocial behavior[31]. It is important to acknowledge that, given the secondary analysis nature of Study 1, the three items stood as our sole available options. To check the robustness of the results, the parental report of prosocial behavior in the follow-up survey was chosen as an alternative measure. The parents were asked to rate the frequency of the same three prosocial behaviors of their children in the past year (1 = never, 5 = always; M = 3.694, SD = 0.786, Cronbach's α = 0.717).

      The independent variable in this study is book reading. It was measured by the average hours the student spent last week reading books (M = 35.097, SD = 37.557). Control variables included two other media use variables (i.e., television and video games), demographic variables (i.e., gender, age, rural hukou, migration, residence in school, parents' education, ethnicity, and single child), cognitive ability variables (i.e., ranking in class, cognitive ability test by CEPS, math test scores, Chinese test scores, and English test scores), family relationship variables (i.e., educational expectations of parents, relationship quality between father and mother, relationship quality with mother, and relationship quality with father), teacher-related variables (i.e., teachers' criticism, teachers' responsibility, teachers' patience, the likability of the student advisor, the likability of other teachers), and peers' relationship variables (i.e., friendliness of classmates, closeness with others in school, and likability of classmates). Class fixed effects were also controlled.

      To further validate the results of Model 1 and 2, we used self-reported and parental reported antisocial behaviors measured in the follow-up survey as dependent variables to examine their relationship with book reading. Such a comparison enables us to check the robustness of the impact of book reading on adolescents' prosocial development. Antisocial behavior was indicated by averaging the frequency of the student's five antisocial behaviors in the past year. The five behaviors were (1) insulting others and talking dirty, (2) quarreling, (3) fighting, (4) bullying classmates, and (5) cheating on homework and exams (1 = never, 5 = always; self-reported: M = 1.609, SD = 0.565, Cronbach's α = 0.762; parental reported: M = 1.366, SD = 0.439, Cronbach's α = 0.747).

    • We used Stata 17 to estimate the OLS models. Table 1 presents the detailed results of the four regression models. Book reading was found to be positively related to both self-reported and parent-reported prosocial behavior (self-reported: B = 0.011, p <0.001; parental reported: B = 0.006, p = 0.009). Therefore, the relationship between book reading and adolescents' prosocial behavior is supported. The relationship is consistent across the combinations of different measures, suggesting such a relationship is not sensitive to measurement. Moreover, the results of Models 3 and 4 revealed that book reading negatively predicts antisocial behavior (self-reported: B = −0.007, p < 0.001; parental reported: B = −0.004, p = 0.002).

      Table 1.  Regression results of study 1.

      VariablesProsocial behaviorAntisocial behavior
      Self-reportedParental reportedSelf-reportedParental reported
      Book reading0.011***0.006**−0.007***−0.004**
      (0.002)(0.002)(0.002)(0.001)
      Media use variables
      Watching TV−0.008***−0.007**0.005***0.004***
      (0.002)(0.002)(0.001)(0.001)
      Playing games−0.005**−0.004ϯ0.014***0.006***
      (0.002)(0.002)(0.001)(0.001)
      Demographic variablesControlledControlledControlledControlled
      Cognitive abilityControlledControlledControlledControlled
      Family relationship variablesControlledControlledControlledControlled
      Teachers related variablesControlledControlledControlledControlled
      Peers' relationship variablesControlledControlledControlledControlled
      Class fixed effectControlledControlledControlledControlled
      Constant2.668***2.309***1.955***1.726***
      (0.224)(0.236)(0.162)(0.128)
      Observations7,4877,3497,4917,361
      R20.1880.1630.2120.192
      Adj. R20.1600.1340.1850.164
      F20.6117.6125.9621.50
      Standard errors in parentheses. ϯ p < 0.1, ** p < 0.01, *** p < 0.001. For the clarity of the table, the coefficients of each type of control variable were not listed in detail.

      More interestingly, the results showed that watching TV is negatively related to prosocial behavior while positively related to antisocial behavior (self-reported prosocial behavior: B = −0.008, p < 0.001; parental reported prosocial behavior: B = −0.007, p = 0.001; self-reported antisocial behavior: B = 0.005, p < 0.001; parental reported antisocial behavior: B = 0.004, p < 0.001). In addition, playing games is negatively related to prosocial behavior (self-reported: B = −0.0005, p < 0.001; parental reported: B = −0.004, p < 0.01) and positively associated with antisocial behavior (self-reported: B = 0.014, p < 0.001; parental reported: B = 0.006, p < 0.001).

    • Based on the analysis of a nationally representative data set, this study confirmed that book reading positively predicted adolescents' prosocial behavior and supported the robustness of such a relationship. The negative relationship between book reading and antisocial behaviors provided additional evidence that book reading is beneficial for adolescents' prosocial development. However, due to the constraints of secondary analysis, this study failed to explore the potential mechanisms underlying the relationship between book reading and adolescents' prosocial behavior. Therefore, we designed a new survey attempting to address this question.

    • This study is based on a survey conducted in an East China city in 2021. The corresponding author's Institutional Review Board (No. H2021177I) approved the protocol. We adapted established scales to develop the questionnaire, which was evaluated by three experienced secondary school teachers for the understandability of adolescents. We then performed a pretest with a convenience sample of 103 adolescents to test all the instruments' reliability and validity. We also modified the questionnaire based on the feedback of the pretest, such as deletion of some items and improvement of the wording.

      We conducted the formal paper-based survey in one junior secondary school and one senior secondary school. Participants were randomly selected from each grade at each school to ensure the distribution of our participants covered all six grades. The number of participants from each grade was approximately one hundred. The initial sample size is 647. We then dropped participants whose responses were over 50% empty or straight-line (i.e., participants who responded to all the 5-point Likert scale with almost the same answer). One participant who indicated age as 100 years old was also removed. Finally, 631 participants were kept (49.18% females and 50.82% males, aged from 12 to 19; Mage = 15.211, SDage = 1.600) and we imput the missing data with the expectation-maximization algorithm for data analysis.

    • Book reading was measured by book reading habits in this study. This measurement is adapted from the measures of social media use by Ellison et al.[32]. Participants were asked to rate their agreement (1 = completely disagree, 5 = completely agree) on the following four statements: (1) 'I often read books in my free time,' (2) 'Reading books in free time is my habit,' (3) 'I've been reading books for years,' (4) 'I feel bad if I don't read books,' and a reading frequency question 'Generally speaking, how frequently do you read books?' (1 = never, 5 = always). The five items were all measured with 5-point Likert scales, and they were averaged for data analysis (M = 3.115, SD = 0.920, Cronbach's α = 0.859).

    • The social norms measure contained three items, which were adapted from the work of Ajzen[33] and Smith & McSweeney[34]. Respondents rated the extent to which they agree or disagree on a 5-point scale (1 = strongly disagree, 5 = strongly agree) with the following statements: 'The majority of people today still have basic moral values,' 'The majority of people will not hesitate to lend a helping hand to people in need,' and 'The majority of people believe that they should treat others the way they would like to be treated.' The scale had a acceptable reliability of α = 0.742 (M = 3.636, SD = 0.855).

    • To measure personal norms, we used the scale modified from scales developed by Smith & McSweeney[34]. Respondents were asked to rate on a 5-point scale (1 = strongly disagree, 5 = strongly agree) the extent to which they agree with three items (M = 3.454, SD = 0.920, Cronbach's α = 0.899). The items included: 'I feel morally obliged to help people in need,' 'I feel guilty when I do not help people in need,' and 'It is my duty to help people in need.'

    • Based on the study of Carlo & Randall[35], Yang et al.[31] developed the Chinese version of the prosocial behavior scale incorporating features of Chinese adolescents. It has been validated and extensively used in measuring adolescents' prosocial behavior in China. The scale contains 15 items. Each item describes a kind of prosocial behavior. Sample items included 'I voluntarily give seats to those in need, such as the elderly, the weak, the sick, the disabled, and the pregnant,' and 'When a classmate is sick, I take him to see the school nurse.' Participants were asked to rate the extent to which the item is in line with their condition (1 = Not at all, 5 = Completely; M = 3.647, SD = 0.785, Cronbach's α = 0.909).

    • Table 2 presents the correlation metrics of variables included in the model. To explore whether social norms and personal norms function as potential mediators linking book reading and adolescents' prosocial behavior, we conducted a serial mediation analysis with PROCESS Model 6 for SPSS[36], in which book reading was entered as the independent variable, prosocial behavior as the dependent variable, social norms as the stage-one mediator, and personal norms as the stage-two mediator. Meanwhile, five variables (e.g., gender, age, parent's education, income, and empathy) were entered into the model as covariates. Table 3 presents the regression results. where social norms, personal norms, and prosocial behaviors were examined as dependent variables, respectively. We examined the indirect effects with biased-corrected confidence interval recommended by Preacher & Hayes[37] (See Table 4). If a confidence interval for the indirect effect does not straddle zero, this can statistically support that M mediates the effect of X on Y[38].

      Table 2.  The correlation metrics of variables.

      MSDPBSNPNRI
      PB3.6470.7851.000
      SN3.6360.8550.5001.000
      PN3.4540.8990.5870.5451.000
      RI3.1150.9200.2600.1670.1841.000
      Note: The p values of all the correlation coefficients < 0.001. PB = prosocial behavior, SN = social norms, PN = personal norms, RI = reading intensity.

      Table 3.  Regression results of study 2.

      Dependent variableSocial normsPersonal normsProsocial behavior
      BSE95% CIBSE95% CIBSE95% CI
      Book reading0.100**0.035[0.031, 0.170]0.059ϯ0.031[−0.002, 0.120]0.113***0.023[0.067, 0.159]
      Social norms0.385***0.035[0.316, 0.453]0.178***0.029[0.122, 0.234]
      Personal norms0.207***0.030[0.148, 0.266]
      Female0.150*0.064[0.025, 0.275]−0.0100.056[−0.119, 0.100]−0.0100.042[−0.092, 0.072]
      Age−0.072***0.019[−0.11, −0.034]−0.057**0.017[−0.091, −0.024]−0.0130.013[−0.039, 0.012]
      Parents' education−0.0160.032[−0.078, 0.047]−0.0090.028[−0.064, 0.045]0.0250.021[−0.016, 0.065]
      Income−0.0680.053[−0.173, 0.036]−0.0640.047[−0.155, 0.028]−0.0180.035[−0.086, 0.051]
      Empathy0.451***0.045[0.363, 0.539]0.516***0.042[0.434, 0.599]0.374***0.035[0.305, 0.442]
      constant2.909***0.387[2.149, 3.669]1.037*0.352[0.345, 1.730]0.675*0.266[1.197, 0.152]
      R20.2000.4520.530
      F25.966***73.264***87.834***
      ϯ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

      Table 4.  Direct and indirect effects.

      BSE95% CI
      Total effect0.1510.025[0.101, 0.201]
      Direct effect0.1130.023[0.067, 0.159]
      Indirect effects
      Total indirect effect0.0380.012[0.015, 0.064]
      Book reading → SN → PB0.0180.007[0.005, 0.033]
      Book reading → PN → PB0.0120.007[0.000, 0.027]
      Book reading → SN → PN → PB0.0080.003[0.002, 0.015]
      Note: The standard error (SE) and 95% CI of indirect effects are based on bias-corrected bootstrap samples. SN = social norms, PN = personal norms, PB = prosocial behavior.

      As for the specific indirect effect through social norms, results showed that book reading was significantly associated with social norms (B = 0.100, p = 0.004, 95% CI = [0.031, 0.170]), which in turn significantly predicted prosocial behavior (B = 0.178, p < 0.001, 95% CI = [0.122, 0.234]). A 95% bias-corrected confidence interval based on 5,000 bootstrap samples indicated that the indirect effect through social norms was entirely above zero (B = 0.018, 95% CI = [0.005, 0.033]). With regard to the indirect effect via personal norms, we found that book reading did not predict personal norms (B = 0.059, p = 0.006, 95% CI = [−0.002, 0.120]), whereas personal norms were significantly and positively associated with prosocial behavior (B = 0.207, p < 0.001, 95% CI = [0.148, 0.266]). A 95% bias-corrected confidence interval based on 5,000 bootstrap samples indicated that there was not an indirect effect through personal norms since it straddled zero (B = 0.012, 95% CI = [0.000, 0.027]). In short, the specific indirect effect of social norms alone existed, whereas that of personal norms alone did not.

      In terms of the indirect effect through social norms and personal norms in serials, book reading significantly predicted social norms (B = 0.100, p = 0.004, 95% CI = [0.031, 0.170]), social norms significantly predicted personal norms (B = 0.385, p < 0.001, 95% CI = [0.316, 0.453]), and personal norms were significantly associated with prosocial behavior (B = 0.207, p < 0.001, 95% CI = [0.148, 0.266]). A 95% bias-corrected confidence interval based on 5,000 bootstrap samples indicated that the indirect effect via social norms and personal norms in serials existed (B = 0.008, 95% CI = [0.002, 0.015]). That is, the sequential mediating effect of social norms and personal norms was found to link book reading and prosocial behavior among adolescents (Fig. 1).

      Figure 1. 

      Model estimation results.

    • Based on the analysis of a self-collected data set. Study 2 explores the relationship between book reading, social norms, personal norms, and prosocial behavior. Through the examination of a serial-mediated model, the study found that the relationship between book reading and prosocial behavior can be mediated by social norms and personal norms. The indirect effect of social norms and the serial mediation of social norms and personal norms are both found to be significant. Such a finding supports the argument that reading books can regulate adolescents' prosocial behavior through social influence. Following, we will discuss the potential reasons for the results, as well as the theoretical and practical contributions, by synthesizing the findings of study 1 and study 2.

    • Through two empirical studies, this research explores the relationship between adolescents' book reading and their prosocial behavior. The first study supports the positive association between book reading and prosocial behavior in adolescents. The second study examines the sequential serial model of social norms and personal norms and finds book reading can predict adolescents' prosocial behavior through the mediation of social norms and personal norms.

      First, this study confirms that reading books positively predicts prosocial behavior among adolescents. Despite the substantial research on the different forms of media use, such as watching TV/films and playing video games, on prosocial behavior (e.g., De Leeuw et al.[12], Greitemeyer[13], Ruth[14], Greitemeyer & Mügge[15]), there are few published findings on the association between book reading as a traditional media use and prosocial behavior. Our research adds to the media effect and prosocial behavior scholarships by focusing on book reading as an often-neglected media use. Besides, prior research often focused on the impact of reading specific book genres or content in books on prosocial behavior[20,39]. Yet, little is known about the effect of book reading in general. This research verifies such a positive effect among adolescents. Therefore, this study and its findings add to the body of knowledge on the prosocial effect of book reading. Practically, knowing the potential positive impacts of reading books on adolescents' prosocial behavior would be valuable in adolescents' moral education. For example, schools and parents can encourage adolescents to read more, which can not only cultivate their reading habits but also help to promote their prosocial development.

      Second, this study shows that the relationship between book reading and prosocial behavior is mediated by social norms alone, as well as social norms and personal norms in serial. On the one hand, the indirect effect through social norms confirmed our argument that adolescents' engagement in book reading leads to their cognitive benefits that drive the prosocial outcome of book reading via the mediation of social norms. Since the previous studies on the prosocial outcomes of media use (including book reading) mostly focus on exposure to specific prosocial contents, intrinsic regulations variables (e.g., moral identity)[20], or social cognitive ability variables (e.g., empathy and prosocial thoughts)[13,16,40,41], and are widely examined. Unlike these more intrinsic variables, the improved perception of social norms, an outcome induced by more frequency of book reading and the increased level of cognitive abilities, involves individuals' perceptions of others' approval and disapproval of media use, which is more extrinsic-oriented. On the other hand, besides the specific indirect effect through social norms alone, there also exists the sequential mediating effect of social norms and personal norms in serial. These results demonstrate the function of social norms in linking book reading and personal norms and confirms the idea of internalizing social norms into personal norms[30].

      Practically speaking, the empirical finding that reading books reinforces prosocial behavior through social and personal norms holds practical implications for educators and parents. Schools can develop curricula with reading books exemplifying positive values, fostering insightful classroom discussions and contributing to the development of empathetic individuals. Simultaneously, parents can establish a nurturing reading environment at home by offering a diverse range of books and encouraging discussions on characters' choices. Collaborative endeavors could include book clubs and experiential activities like role-playing. By integrating these strategies, schools and parents can jointly nurture well-rounded individuals who excel academically and exhibit empathy, compassion, and social responsibility in their interactions with others and society at large.

      Finally, several limitations should be noted regarding this study. First, the two studies are based on surveys, which are limited in inferring causal relationships. Future studies can adopt experiments to establish the causal relationships between book reading and prosocial behaviors. Second, although the first study is based on nationally representative data, the data of the second study is collected in one province of China. Thus, the generalizability of the second study is relatively limited. Future studies could seek to collect more representative data and do cross-country or cultural studies to increase the generalizability of the findings. Additionally, we want to recognize that there exist other potential factors that might influence the relationships, such as certain personality traits and the genres of books. Although these aspects are not the central focus of our present research, they could serve as valuable directions for future investigations. Finally, we found that only reading books can positively predict prosocial behavior, while watching television and playing video games cannot exert a similar effect. Future studies can focus on the potential different effects of various media types on adolescents' prosocial behavior.

      • The authors declare that they have no conflict of interest.

      • Copyright: © 2023 by the author(s). Published by Maximum Academic Press, Fayetteville, GA. This article is an open access article distributed under Creative Commons Attribution License (CC BY 4.0), visit https://creativecommons.org/licenses/by/4.0/.
    Figure (1)  Table (4) References (41)
  • About this article
    Cite this article
    Ai P, Li W. 2023. Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior. Publishing Research 2:5 doi: 10.48130/PR-2023-0005
    Ai P, Li W. 2023. Does reading increase prosociality? Linking book reading with adolescents' prosocial behavior. Publishing Research 2:5 doi: 10.48130/PR-2023-0005

Catalog

  • About this article

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return