Water self-diffusion differentiates two protein-based burger matrices with contrasting sensory juiciness Authors: Moshe H Azachi, Zeev Wiesman Published: 2026 ABSTRACT AbstractJuiciness is a major sensory attribute of meat and plant-based protein foods, but bulk moisture content alone does not necessarily describe how water is organized within a food matrix. This study compared two structurally distinct commercial burger formulations—Angus beef and soy-based burgers—to determine whether time-domain nuclear magnetic resonance (TD-NMR) measurements of the pre-deformation water state distinguish matrices with contrasting sensory juiciness. Cooked burgers were characterized by TD-NMR relaxometry and apparent water self-diffusion, Cryo-SEM, texture profile analysis, cooking measurements, and trained sensory evaluation. At the independent parent-burger level, apparent water self-diffusion, D, was higher in beef than in soy-based burgers (1.030 ± 0.067 vs. 0.704 ± 0.070 × 10−9 m2 s−1; n = 5 per formulation; p < 0.001), whereas mean long-T2 did not differ significantly (p = 0.163). The exploratory D/long-T2 ratio was also higher in beef (p = 0.026). Trained assessors rated beef as juicier than the soy-based burger (SJI 7.46 ± 1.41 vs. 5.00 ± 1.00; n = 13 paired assessors; p < 0.001), while initial moisture content did not differ significantly between formulations (p = 0.064). Cryo-SEM showed qualitative formulation-dependent differences in matrix morphology. TD-NMR and sensory measurements were obtained from separate experimental units; therefore, these results demonstrate parallel formulation-level differences and do not establish a within-sample correlation between D and sensory juiciness. Moreover, molecular self-diffusion is physically distinct from deformation-induced serum flow during mastication. The data support apparent water self-diffusion as a sensitive descriptor of the pre-deformation water state of these two matrices and motivate future studies coupling TD-NMR, mechanical serum release, and sensory measurements within the same independent samples.Keywords: Time-domain NMR, Water self-diffusion, Sensory juiciness, Plant-based burger, Food microstructure, Water mobility Introduction Juiciness is one of the most influential sensory attributes determining consumer acceptance of conventional meat products and plant-based meat analogues (PBMAs). It contributes not only to perceived moistness but also to flavor release, lubrication, ease of oral processing, and overall eating quality. Although considerable progress has been achieved in plant-protein structuring, extrusion, fat structuring, and ingredient formulation, reproduction of the characteristic juiciness of animal-derived meat remains a major technological challenge for PBMAs (Dekkers et al., 2018; Kyriakopoulou et al., 2019; Samard et al., 2019; Sha et al., 2020; He et al., 2020; Ismail et al., 2020; Zhang et al., 2024a, 2024b, 2026). Spatial heterogeneity in food matrices can also modulate sensory perception, underscoring the importance of structure rather than composition alone (Mosca et al., 2010). Juiciness has traditionally been related to moisture content, water-holding capacity, expressible moisture, and cooking loss (Aaslyng et al., 2003; Dekkers et al., 2016). These measurements provide important information about the quantity of water retained within a product but do not necessarily indicate how much fluid becomes available during mastication. Food oral processing is dynamic: deformation, fracture, saliva incorporation, bolus formation, fluid release, and lubrication occur simultaneously and evolve throughout chewing (Chen, 2009; Wilkinson et al., 2000). Consequently, a product can contain substantial quantities of water and nevertheless be perceived as dry if that water remains strongly confined or is poorly released during oral deformation. Recent studies have considerably refined this understanding. Zhang et al., 2024a, 2024b demonstrated strong relationships between serum release during mastication and dynamic juiciness perception in meat-analogue and beef patties. More recent work showed that the amount of serum released can be more important for juiciness perception than the composition or viscosity of the serum itself (Zhang et al., 2026). Rudge et al. (2025) similarly emphasized the importance of measurement conditions when relating instrumental water release to sensory juiciness. Together, these studies identify mechanically induced serum availability during oral processing as a central proximal determinant of perceived juiciness. An important mechanistic question nevertheless remains unresolved. Serum release during mastication is a deformation-driven process. Compression and fracture of a hydrated soft material generate pressure gradients that can displace fluid through interconnected pathways. This macroscopically driven transport is physically different from molecular self-diffusion. Water self-diffusion measured by TD-NMR describes spontaneous molecular translational motion in the absence of the imposed macroscopic pressure gradients characteristic of mastication. Consequently, a self-diffusion coefficient should not be interpreted as the velocity or rate of serum expulsion during chewing. The two processes may nevertheless share structural determinants. Network dimensions, connectivity of aqueous domains, tortuosity, interfaces, water-protein or water-polysaccharide interactions, and dispersed lipid phases can influence molecular translational mobility and can also influence the resistance experienced by fluid when a pressure gradient is subsequently imposed. Confined-compression experiments on soy protein gels and model meat analogues provide direct evidence that mechanically induced water-release kinetics depend on material structure and that internal water-filled cavities can provide pathways of relatively low hydraulic resistance (Cornet et al., 2020). More generally, network mesh dimensions can modulate diffusion through hydrated polymer matrices (Liu et al., 2026). These observations suggest that measurements of the pre-deformation molecular water state may contain information relevant to the architecture that subsequently controls deformation-induced fluid redistribution. Food microstructure is therefore central to understanding juiciness (Aguilera, 2005; Wiesman, 2026; Xue et al., 2025a, 2025b; Tornberg, 2005). Protein-network organization, pore-like domains, interfaces, lipid distribution, hydrocolloid organization, and structural anisotropy can modify both local water environments and longer-range molecular mobility. Xue et al. (2025a) demonstrated relationships between structural pores and the sensory and mechanical properties of plant-based burgers, while changes in oil distribution and stability can alter textured vegetable protein hydration and juiciness (Xue et al., 2025b). In meat, heating similarly reorganizes myofibrillar structures and water-containing compartments, thereby changing both water distribution and mechanical properties (Tornberg, 2005; Bertram and Andersen, 2007). Hydration level and particle size have likewise been shown to modify juiciness and flavor perception in meat-analogue patties (Zhang et al., 2025). Low-field time-domain nuclear magnetic resonance (TD-NMR) provides a non-destructive means of probing these water states. Transverse relaxation, T2, is sensitive to molecular environments, exchange, interactions with macromolecules, and confinement, whereas pulsed-field-gradient measurements provide an apparent water self-diffusion coefficient reflecting translational molecular displacement over the experimental observation scale (Bertram and Andersen, 2007; Todt et al., 2006; Mariette, 2009; van et al., 2010; Hills, 2017; Kirtil et al., 2017; Colnago et al., 2021; Acri et al., 2021; Guo et al., 2023). These measurements therefore provide information that differs fundamentally from gravimetric moisture determination. Previous work from our laboratory demonstrated that animal-derived and plant-based burger matrices exhibit different TD-NMR relaxation characteristics, self-diffusion behavior, texture development, and microstructural organization (Wiesman, 2026; Wiesman et al., 2025; Azachi and Wiesman, 2026a; Azachi and Wiesman, 2026b). However, the relationship between TD-NMR-derived water mobility and directly reported sensory juiciness remained to be established. For clarity, water accessibility in the present study is used as an operational structural concept rather than a directly measured quantity. It refers to the degree to which retained water occupies molecular and microstructural environments that permit translational motion within the constraints of the food matrix. It does not imply that TD-NMR directly measures mechanically extractable serum or serum-release kinetics. The specific novelty of the present study is the formulation-level comparison of TD-NMR translational water mobility with sensory juiciness while explicitly separating molecular self-diffusion from mechanically induced serum release. The objectives were to: (i) compare water relaxation and self-diffusion behavior in cooked Angus beef and soy-based burger matrices; (ii) compare these measurements with microstructure, texture, cooking behavior, and sensory juiciness at the formulation level; and (iii) determine whether apparent water self-diffusion provided discrimination between the two matrices beyond that provided by bulk moisture content. An exploratory Water Accessibility Index (WAI = D/T2) was additionally evaluated as an empirical combination of diffusion and relaxation behavior. The study was not designed to establish a universal beef-versus-plant distinction or a causal relationship between D and juiciness. Because sensory and TD-NMR measurements were not obtained as matched observations from the same parent burgers, no within-sample correlation between D and SJI was calculated or inferred. Materials and methods Materials Two commercially available burger systems representing structurally distinct protein-based food matrices were investigated. Angus beef burgers (Black Angus beef; Loup River Beef, Nebraska, USA) were selected as the animal-derived protein system. Soy-based burgers (Tivall Vegan Soybean Burger; Osem-Nestlé, Israel) were selected as the plant-based system and contained soy proteins, vegetable oils, and hydrocolloid ingredients. The purpose was not to establish universal differences between animal- and plant-derived burgers or to compare commercial brand performance. The products served as two contrasting matrices with different compositions and structural organizations; conclusions are therefore restricted to the formulations investigated. Burger preparation and cooking procedure Burgers were cooked in an air fryer preheated to 200 °C until the geometric center reached 72 °C, monitored with a calibrated digital thermocouple. Samples were equilibrated for 5 min at ambient temperature before analysis. For spatial assessment, cooked samples were separated into an outer region approximately 3-5 mm below the surface and an interior core region. These regions were evaluated separately where applicable by TD-NMR and Cryo-SEM. Independently prepared whole burgers constituted the experimental units. Interior and outer pieces and repeated measurements taken from the same burger were treated as subsamples or technical observations and were not counted as independent samples. Moisture content and cooking characteristics Moisture content was determined gravimetrically according to AOAC International procedures (AOAC International, 2019). Approximately 5 g of homogenized sample was dried at 105 ± 2 °C to constant mass, and moisture was expressed as percentage water relative to sample mass. where Wᵢ and Wf are burger masses before and after cooking, respectively (Aaslyng et al., 2003). where Dᵣ and Dc are burger diameters before and after cooking, respectively (Serdaroğlu et al., 2005). Raw moisture measurements were based on independent burgers, and cooking characteristics were determined from independently prepared burgers; technical measurements from the same burger were averaged before inferential analysis. TD-NMR relaxometry Low-field TD-NMR measurements were performed using a Bruker Minispec mq20 spectrometer (Bruker BioSpin GmbH, Germany) operating at a proton resonance frequency of 20 MHz. Measurements were performed at 40 ± 1 °C. Transverse relaxation was acquired using a Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence (Carr et al., 1954; Meiboom and Gill, 1958) with a recycle delay of 5 s, echo time of 0.2 ms, 8000 echoes, and 16 scans. Relaxation decays were analyzed by inverse Laplace transformation with Tikhonov regularization using CONTIN. Shorter-T2 components were interpreted as more restricted proton environments, whereas longer-T2 components represented less restricted environments. These assignments are operational and do not imply unique microscopic water compartments. Water self-diffusion measurements Water self-diffusion coefficients (D) were determined using pulsed-field-gradient TD-NMR. Signal attenuation was analyzed according to the Stejskal-Tanner relationship (Tanner et al., 1968): where I is signal intensity at gradient strength g, I0 is the initial signal intensity, γ is the proton gyromagnetic ratio, δ is gradient-pulse duration, Δ is diffusion observation time, and D is the apparent self-diffusion coefficient. D is interpreted as a molecular-scale descriptor of translational water mobility under the measurement conditions; it is not a macroscopic flow velocity, permeability coefficient, or direct measure of serum-release rate. For cooked diffusion analysis, technical/subsample measurements from each independently prepared parent burger were averaged before group-level inference. Exploratory Water Accessibility Index WAI was retained as an exploratory empirical comparative descriptor combining two experimentally distinct aspects of water behavior. It is unit-dependent and is not interpreted as hydraulic permeability, serum-release rate, or a validated material constant. Cryogenic scanning electron microscopy Cooked burger microstructure was characterized by Cryo-SEM using established principles for hydrated materials (Donald, 2003). Small sections from interior and outer regions were cryo-immobilized by high-pressure freezing (EM ICE, Leica Microsystems, Vienna, Austria), transferred under cryogenic conditions using an EM VCT500 system, fractured in an EM ACE900 preparation system, etched at −100 °C for 5 min, and sputter-coated with approximately 3 nm platinum. Images were acquired using a Gemini scanning electron microscope (Zeiss, Oberkochen, Germany) with an in-lens secondary-electron detector at 5-10 kV and specimen temperature of approximately −120 °C. Cryo-SEM was used qualitatively; pore size, mesh size, tortuosity, and three-dimensional connectivity were not quantified. Texture profile analysis Texture profile analysis (TPA) was performed using a double-compression procedure consistent with Bourne (2002). Standardized cylindrical samples were compressed twice to 50% strain using a texture analyzer fitted with a flat cylindrical probe. Hardness, cohesiveness, springiness, and chewiness were obtained from the resulting force-time curves. Three independently prepared parent burgers per formulation were analyzed, with three technical/subsample measurements averaged within each parent burger before inferential comparison. No composite TPA index was used in the final analysis. Sensory evaluation and Sensory Juiciness Index Sensory evaluation was conducted with 13 trained assessors using a structured 10-point intensity scale, consistent with established meat sensory-evaluation guidance (American Meat Science Association, 2015). Five juiciness-related attributes were evaluated: first-bite juiciness (FB), sustained juiciness during chewing (SJ), perceived rate of juice release (RR), mouthcoating/oral wetness (MC), and perceived dryness (DR). Samples were coded and presented under controlled conditions according to the laboratory sensory protocol. Dryness was reverse-scored, and the composite Sensory Juiciness Index (SJI) was calculated as: Higher SJI represents greater integrated perceived juiciness. The sensory work involved adult trained assessors who participated voluntarily and provided informed consent in accordance with the applicable institutional requirements for sensory evaluation. Statistical analysis Results are expressed as mean ± standard deviation (SD). The independently prepared parent burger was the experimental unit for physicochemical and instrumental measurements; technical or subsample measurements from the same burger were averaged before inferential analysis. Two-formulation comparisons were performed using two-sided Welch independent-samples t-tests where the burger was the experimental unit. Sensory comparisons used paired two-sided t-tests because the same 13 trained assessors evaluated both formulations. For regional TD-NMR comparisons involving more than two groups, analysis of variance followed, where applicable, by Tukey's honestly significant difference test was used. Statistical significance was set at p < 0.05. Because sensory ratings and TD-NMR measurements were not obtained as matched observations from the same parent burgers, Pearson correlations between SJI and TD-NMR, cooking, texture, or moisture variables were not calculated in the final analysis. Statistical analysis and graphical visualization were performed using OriginPro (OriginLab Corporation, Northampton, MA, USA). Results Composition and cooking behavior differed markedly between matrices Initial composition and cooking characteristics are summarized in Table 1. Initial moisture content was 68.05 ± 3.46% in Angus beef burgers and 62.61 ± 5.25% in soy-based burgers; this difference did not reach statistical significance (p = 0.064). After cooking, beef burgers underwent substantially greater cooking loss (35.73 ± 3.62% vs. 21.21 ± 3.75%; p < 0.001) and dimensional shrinkage (39.25 ± 7.30% vs. 17.90 ± 6.88%; p < 0.001) than soy-based burgers. Thus, the soy-based matrix retained mass and dimensional stability more effectively during cooking but, as shown below, was perceived as substantially less juicy. Table 1Initial physicochemical composition and cooking performance.Table 1ParameterAngus beef burgerSoy-based burgerp-valueInitial moisture content (%)68.05 ± 3.4662.61 ± 5.250.064Cooking loss (%)35.73 ± 3.6221.21 ± 3.75< 0.001Shrinkage (%)39.25 ± 7.3017.90 ± 6.88< 0.001Values are mean ± SD. Initial moisture: n = 6 independent burgers per formulation; cooking loss: n = 8; shrinkage: n = 6. P-values are from two-sided Welch independent-samples t-tests. Initial physicochemical composition and cooking performance. Values are mean ± SD. Initial moisture: n = 6 independent burgers per formulation; cooking loss: n = 8; shrinkage: n = 6. P-values are from two-sided Welch independent-samples t-tests. TD-NMR relaxation revealed distinct water environments Multi-exponential T2 analysis resolved four principal relaxation populations in both matrices (Table 2; Fig. 1). Beef burgers generally exhibited longer T2 values for the shorter and intermediate components. In the interior, T22 and T23 were 61.07 ± 6.03 and 162.85 ± 28.35 ms in beef compared with 33.00 ± 4.14 and 132.27 ± 28.22 ms in the soy-based matrix. The longest T2 component exceeded approximately 300 ms in all groups and did not differ significantly. Thus, the presence of a relatively mobile local water population alone did not establish how freely water underwent translational displacement through the heterogeneous matrix. Table 2TD-NMR T2 relaxation characteristics of cooked burger regions.Table 2SampleRegionT21 (ms)T22 (ms)T23 (ms)T24 (ms)Mean long-T2 (ms)BeefInterior18.73 ± 1.75a61.07 ± 6.03a162.85 ± 28.35a360.60 ± 79.82a261.7BeefOuter14.33 ± 1.63b54.87 ± 7.70a149.79 ± 22.44a346.57 ± 56.91a248.2Soy-basedInterior11.20 ± 0.77c33.00 ± 4.14b132.27 ± 28.22b375.15 ± 48.99a253.7Soy-basedOuter9.47 ± 1.30d41.07 ± 13.41b133.00 ± 28.12b339.20 ± 82.09a236.1Values are mean ± SD. Technical/subsample measurements from the same independently prepared burger were aggregated before inferential analysis. Different superscripts within a column indicate significant differences (p < 0.05). TD-NMR T2 relaxation characteristics of cooked burger regions. Values are mean ± SD. Technical/subsample measurements from the same independently prepared burger were aggregated before inferential analysis. Different superscripts within a column indicate significant differences (p < 0.05). TD-NMR T2 relaxation characteristics of cooked Angus beef and soy-based burgers. Representative relaxation distributions and principal relaxation populations demonstrate formulation- and region-dependent differences in local proton environments. Assignments indicate relative restriction rather than uniquely defined microscopic water compartments. Water self-diffusion strongly differentiated the burger matrices The apparent water self-diffusion coefficient showed pronounced formulation- and region-dependent differences (Table 3; Fig. 2). The highest D occurred in beef interior (1.03 ± 0.08 × 10−9 m2 s−1), whereas the lowest value occurred in soy-based outer regions (0.36 ± 0.11 × 10−9 m2 s−1). Beef outer and soy-based interior regions showed intermediate values of 0.75 ± 0.08 and 0.70 ± 0.09 × 10−9 m2 s−1, respectively. In both formulations, D decreased from the interior toward the cooked surface. These data describe molecular translational displacement in the undeformed sample and are not direct measurements of macroscopic water-release rate. Table 3Water self-diffusion, mono-exponential T2, and exploratory WAI.Table 3SampleRegionD (×10−9 m2 s−1)Mono-T2 (ms)D/T2BeefInterior1.03 ± 0.08a57.93 ± 7.68b0.018aBeefOuter0.75 ± 0.08b59.74 ± 9.11b0.013bSoy-basedInterior0.70 ± 0.09b61.94 ± 18.83b0.012bSoy-basedOuter0.36 ± 0.11c122.82 ± 30.82a0.003cWAI is an empirical, unit-dependent comparative descriptor and is not interpreted as hydraulic permeability or serum-release rate. Technical/subsample measurements from the same parent burger were aggregated before group-level inference. Water self-diffusion, mono-exponential T2, and exploratory WAI. WAI is an empirical, unit-dependent comparative descriptor and is not interpreted as hydraulic permeability or serum-release rate. Technical/subsample measurements from the same parent burger were aggregated before group-level inference. TD-NMR water self-diffusion and relaxation measurements in cooked Angus beef and soy-based burgers. D represents molecular translational water mobility under the NMR measurement conditions and not mechanically induced serum-flow velocity. The exploratory D/T2 ratio provided complementary information WAI followed the ordering Beef interior > Beef outer ≈ Soy interior ≫ Soy outer. The soy outer layer combined a low apparent diffusion coefficient with a comparatively long mono-exponential T2, illustrating that relatively long local relaxation does not necessarily correspond to high translational mobility through a heterogeneous matrix. Because D/T2 is unit-dependent and lacks a direct theoretical mapping onto hydraulic permeability or serum release, WAI is considered exploratory. Across the two formulations, D provided a clearer instrumental contrast than mean long-T2; no D–SJI correlation was calculated because the TD-NMR and sensory observations were not matched within the same parent burgers. Cryo-SEM revealed contrasting matrix architectures Representative Cryo-SEM images demonstrated clear morphological differences between the two cooked burger systems (Fig. 3). Beef samples showed heterogeneous protein-fat structures containing irregular pore-like and hydrated domains with apparently greater continuity. The soy-based matrix appeared denser and more homogeneous, with thicker solid regions and lower apparent connectivity, particularly in the cooked outer layer. These observations paralleled the TD-NMR measurements but do not establish quantitative pore or network mesh sizes, three-dimensional connectivity, tortuosity, hydraulic permeability, or water-matrix affinity. Representative Cryo-SEM microstructure of cooked Angus beef and soy-based burgers from outer (OUT) and interior (IN) regions. Images illustrate formulation- and region-dependent differences in matrix morphology. Structural descriptions are qualitative; pore size, mesh size, connectivity, and tortuosity were not quantified. Scale bars = 500 nm; inset scale bars = 200 nm. Mechanical properties differed between matrices At the parent-burger level, beef burgers showed numerically greater hardness and chewiness than soy-based burgers, but these differences were not statistically significant with the available independent replication (n = 3 parent burgers per formulation; Table 4). Cohesiveness was modestly but significantly higher in beef (p = 0.033), whereas springiness did not differ. These results indicate that the sensory juiciness contrast cannot be reduced to a simple soft-versus-hard distinction. Table 4Texture profile analysis of cooked burgers.Table 4Burger systemHardness (N)CohesivenessSpringinessChewiness (N·mm)Angus beef24.36 ± 6.950.190 ± 0.007a0.241 ± 0.0021.12 ± 0.34Soy-based12.51 ± 3.610.172 ± 0.007b0.242 ± 0.0020.53 ± 0.17Values are mean ± SD for n = 3 independently prepared parent burgers per formulation after averaging three technical/subsample measurements within each burger. Welch-test p-values: hardness = 0.079; cohesiveness = 0.033; springiness = 0.519; chewiness = 0.073. Different superscripts indicate a significant formulation difference (p < 0.05). Texture profile analysis of cooked burgers. Values are mean ± SD for n = 3 independently prepared parent burgers per formulation after averaging three technical/subsample measurements within each burger. Welch-test p-values: hardness = 0.079; cohesiveness = 0.033; springiness = 0.519; chewiness = 0.073. Different superscripts indicate a significant formulation difference (p < 0.05). Sensory juiciness was substantially higher in beef burgers Sensory evaluation showed significant paired differences across all five juiciness-related attributes (Table 5). Beef burgers received higher ratings for first-bite juiciness, sustained juiciness, perceived rate of juice release, and mouthcoating, whereas soy-based burgers were perceived as drier. The composite SJI, calculated directly from the five sensory attributes for each assessor, was 7.46 ± 1.41 for beef and 5.00 ± 1.00 for soy-based burgers (n = 13 paired assessors; p < 0.001). Table 5Sensory juiciness attributes.Table 5Sensory attributeAngus beefSoy-basedFirst-bite juiciness7.92 ± 1.50a4.77 ± 2.01bSustained juiciness7.62 ± 2.14a4.85 ± 1.46bRate of juice release6.69 ± 2.18a4.92 ± 1.55bMouthcoating/oral wetness7.62 ± 1.71a5.08 ± 1.85bPerceived dryness2.54 ± 2.37b4.62 ± 2.63aSJI7.46 ± 1.41a5.00 ± 1.00bValues are mean ± SD for 13 trained assessors. Formulations were compared within assessor using paired two-sided t-tests. All five sensory attributes and SJI differed significantly between formulations (p < 0.05); SJI p = 0.00039. Different superscripts indicate significant differences. Sensory juiciness attributes. Values are mean ± SD for 13 trained assessors. Formulations were compared within assessor using paired two-sided t-tests. All five sensory attributes and SJI differed significantly between formulations (p < 0.05); SJI p = 0.00039. Different superscripts indicate significant differences. Water self-diffusion most clearly differentiated matrices with contrasting sensory juiciness Among the TD-NMR descriptors examined at the independent parent-burger level, water self-diffusion showed the largest and most statistically robust formulation difference (Table 6; Fig. 4). Mean D was 1.030 ± 0.067 × 10−9 m2 s−1 in Angus beef and 0.704 ± 0.070 × 10−9 m2 s−1 in soy-based burgers (n = 5 parent burgers per formulation; p < 0.001). Mean long-T2 did not differ significantly (267.07 ± 29.21 vs. 240.67 ± 24.81 ms; p = 0.163), whereas the exploratory WAI was higher in beef (0.00396 ± 0.00048 vs. 0.00313 ± 0.00049; p = 0.026). In parallel, the paired sensory panel showed higher SJI for beef (7.46 ± 1.41) than soy-based burgers (5.00 ± 1.00; p < 0.001). These are concordant formulation-level differences. Because the sensory and TD-NMR measurements were not matched within the same parent burgers, a within-sample Pearson correlation between D and SJI cannot be estimated from this dataset. Table 6Formulation-level comparison of principal TD-NMR water-state descriptors and Sensory Juiciness Index.Table 6ParameterAngus beefSoy-basedp-valueWater self-diffusion coefficient, D (×10−9 m2 s−1)1.030 ± 0.0670.704 ± 0.070< 0.001Mean long-T2 (ms)267.07 ± 29.21240.67 ± 24.810.163Exploratory WAI (D/long-T2)0.00396 ± 0.000480.00313 ± 0.000490.026Sensory Juiciness Index (SJI)7.46 ± 1.415.00 ± 1.00< 0.001TD-NMR values are mean ± SD at the parent-burger level (n = 5 independently prepared burgers per formulation), after averaging three technical/subsample measurements within each parent burger. SJI values are mean ± SD for 13 trained assessors who evaluated both formulations and were compared using a paired t-test. TD-NMR comparisons used Welch independent-samples t-tests. WAI is an exploratory, unit-dependent descriptor. Sensory and TD-NMR observations were not matched within the same parent burgers; therefore, no within-sample correlation was calculated. Formulation-level comparison of principal TD-NMR water-state descriptors and Sensory Juiciness Index. TD-NMR values are mean ± SD at the parent-burger level (n = 5 independently prepared burgers per formulation), after averaging three technical/subsample measurements within each parent burger. SJI values are mean ± SD for 13 trained assessors who evaluated both formulations and were compared using a paired t-test. TD-NMR comparisons used Welch independent-samples t-tests. WAI is an exploratory, unit-dependent descriptor. Sensory and TD-NMR observations were not matched within the same parent burgers; therefore, no within-sample correlation was calculated. Formulation-level comparison of TD-NMR water self-diffusion and Sensory Juiciness Index (SJI) in cooked Angus beef and soy-based burgers. TD-NMR D values represent n = 5 independently prepared parent burgers per formulation after averaging technical/subsample measurements within each burger; SJI values represent n = 13 paired trained assessors. D and SJI were obtained from separate experimental units and are therefore presented as parallel formulation-level contrasts rather than as a within-sample correlation. D represents molecular translational water mobility under the NMR measurement conditions and not mechanically induced serum-flow velocity. Evidence-separated multiscale framework Integration of the results supports a multiscale interpretation rather than a direct molecular-diffusion-to-juiciness mechanism (Fig. 5). At the molecular scale, TD-NMR directly demonstrated different translational water mobility and relaxation behavior between the matrices. At the sensory scale, the trained panel directly demonstrated different perceived juiciness. These observations were obtained from separate experimental units and therefore establish concordant formulation-level contrasts, not a within-burger correlation. The proposed intermediate pathway—architecture-dependent constraints on water, deformation-induced pressure gradients, macroscopic serum displacement, and oral lubrication—remains a mechanistic hypothesis requiring direct validation. Proposed multiscale architecture-water-state-juiciness framework. Food architecture defines the physical environment experienced by water, while TD-NMR directly measures molecular relaxation and self-diffusion in the undeformed matrix. Network dimensions, aqueous-domain connectivity, tortuosity, and water-matrix interactions are proposed structural determinants of this molecular state. During mastication, deformation generates pressure gradients that may drive macroscopic serum displacement through available pathways. Serum release and oral lubrication were not measured in the present study and are hypothesized intermediates. Sensory juiciness was measured directly. Solid elements denote measurements or formulation-level observations supported by the present data; dashed elements denote proposed mechanisms requiring independent validation. Discussion Water quantity did not explain the sensory contrast A central observation was the mismatch between bulk water retention and sensory juiciness. The soy-based burgers experienced markedly less cooking loss yet were perceived as substantially less juicy than the beef burgers. Initial moisture content itself did not differ significantly between formulations (p = 0.064). These results reinforce that bulk water quantity and retention during cooking are insufficient, by themselves, to account for the sensory contrast between the two matrices. Molecular self-diffusion and mechanically induced serum flow are distinct phenomena Molecular self-diffusion measured by TD-NMR and serum flow produced by compression or mastication cannot be treated as equivalent transport processes. Self-diffusion reflects thermally driven molecular translational motion and can be influenced by obstacles, confinement, exchange, interfaces, pathway geometry, and the observation scale of the NMR experiment (Bertram and Andersen, 2007; van et al., 2010; Laghi et al., 2017; Tanner et al., 1968). Serum release during mastication is different: deformation generates local stresses and pressure gradients, and fluid displacement depends on permeability, mechanics, fracture, drainage length, and connectivity of fluid-containing domains. Therefore, D is not serum-flow velocity and should not be interpreted as a rate constant for juice release. The proposed link is architectural. Some structural properties that restrict molecular displacement may also increase resistance to deformation-induced fluid movement. Cornet et al. (2020) showed by confined compression that mechanically induced water-release kinetics in soy protein gels and model meat analogues depend strongly on internal structure. In the present dataset, higher D and higher sensory juiciness occurred in the same formulation, but the measurements were not paired within the same parent burgers. This concordance suggests that D may capture information about the pre-deformation water-containing architecture relevant to eating quality, while direct causal and within-sample relationships remain to be tested. Why was water self-diffusion higher in the beef matrix? Several non-exclusive mechanisms may contribute to the higher D measured in the beef matrix. First, characteristic dimensions of water-containing domains or network meshes may differ; decreasing accessible dimensions can reduce effective diffusivity in hydrated polymer networks (Liu et al., 2026). Second, connectivity and tortuosity may differ: water can remain locally mobile yet encounter barriers to longer-range displacement. Third, water-matrix affinity may contribute. Soy-based formulations contain proteins, hydrocolloids, and other hydrophilic components capable of modifying water residence, exchange, and confinement; phase-specific water redistribution in plant-protein blends has been demonstrated by TD-NMR (Schreuders et al., 2020). Fourth, dispersed phases and interfaces can increase obstruction. Conversely, cooked muscle retains hierarchical organization derived from fibers, myofibrillar structures, extracellular spaces, fat domains, and cooking-induced voids. The Cryo-SEM images were consistent with a more heterogeneous and apparently connected beef architecture. However, mesh size, three-dimensional connectivity, tortuosity, permeability, and water-matrix interaction energy were not directly quantified, so these explanations remain mechanistic hypotheses rather than demonstrated causes. T2 and D provide complementary information The results illustrate why relaxation and diffusion should not be treated interchangeably. T2 is influenced primarily by local molecular environments, interactions, exchange, and effective restriction, whereas PFG-derived D reflects translational displacement over the diffusion experiment's spatial and temporal window. The longest relaxation components overlapped substantially between formulations, whereas D differed markedly. The soy-based outer layer was particularly informative: mono-exponential T2 remained relatively long while D was very low. Local mobility can therefore remain appreciable even when longer-range translational displacement is restricted. Neither measurement alone is equivalent to mechanical serum release. These distinctions are consistent with established low-field NMR interpretations of water mobility and relaxation in heterogeneous foods (Bertram et al., 2002; Hills, 2006). Microstructure provides context for the diffusion differences Cryo-SEM demonstrated clear qualitative differences in cooked matrix morphology. Beef showed a more heterogeneous architecture containing irregular void-like and hydrated regions, whereas the soy-based burger displayed a denser appearance and especially strong consolidation in the cooked outer region. These observations were consistent with the spatial ordering of D. However, Cryo-SEM provides static two-dimensional views and does not directly determine the three-dimensional network through which water moves. The appropriate conclusion is therefore one of consistency across measurements rather than direct proof of a particular pore or mesh mechanism. Relation to the established serum-release theory of juiciness Zhang et al., 2024a, 2024b, 2026 established that serum released during oral processing is a strong proximal predictor of juiciness. The present study addresses an upstream question: whether the physical state of water before deformation is associated with subsequent sensory performance. A multiscale sequence can therefore be proposed: food architecture → pre-deformation molecular water state → deformation-induced fluid redistribution → serum availability/lubrication → juiciness. Only selected elements were measured here. Architecture was examined qualitatively by Cryo-SEM; molecular water state was measured by TD-NMR; and sensory juiciness was measured by the trained panel. Deformation-induced serum flux, permeability, and oral lubrication were not measured. The intermediate portion of the sequence therefore remains a testable hypothesis. Future work should measure D before deformation, hydraulic or poroelastic water release during controlled compression, actual serum release during mastication, and sensory juiciness in the same independent samples. Interpretation of WAI should remain exploratory The empirical D/T2 ratio was introduced to combine translational mobility with relaxation behavior. At the parent-burger level, WAI was higher in beef than in soy-based burgers (p = 0.026), whereas mean long-T2 did not differ significantly. Because D/T2 is unit-dependent and has no derived physical relationship to permeability, serum flux, or deformation-driven transport, WAI is retained only as an exploratory within-dataset descriptor. The present data do not establish WAI as a predictor of sensory juiciness because sensory and TD-NMR measurements were not matched within the same parent burgers. External validation across formulations, instruments, field strengths, temperatures, diffusion times, and processing conditions would be required before WAI could be proposed as a transferable material index. Mechanical texture alone did not explain juiciness Beef burgers showed numerically greater hardness and chewiness than the soy-based burgers, although these differences were not statistically significant when technical measurements were aggregated to the three independent parent burgers per formulation. Cohesiveness was modestly higher in beef. Thus, the observed juiciness contrast cannot be explained simply by instrumental hardness or chewiness. Mechanical properties nevertheless remain relevant because they control how a matrix deforms, fractures, and generates pressure gradients during mastication. A complete physical model of juiciness would ideally combine network mechanics, permeability or drainage behavior, water state, fracture, and oral lubrication. Implications for plant-based food design Formulation strategies aimed only at maximizing water-holding capacity or minimizing cooking loss may be insufficient. A highly hydrated matrix may still provide poor juiciness if its architecture strongly restricts redistribution of fluid during oral deformation. Potential design variables include protein-network organization, domain connectivity, structural anisotropy, lipid distribution, protein-polysaccharide interactions, cross-linking, and dimensions and connectivity of hydrated domains. The design objective is not simply to maximize diffusion, but to retain water during manufacture and cooking while making an appropriate fraction available when the product is mechanically deformed during eating. TD-NMR diffusion may be useful as a rapid, non-destructive screening measurement of this pre-deformation water state. Limitations and future work Only one commercial beef formulation and one commercial soy-based formulation were studied; the results therefore establish a contrast between two matrices rather than a universal animal-versus-plant distinction. Molecular self-diffusion was measured, but hydraulic permeability and deformation-induced serum flux were not. Cryo-SEM provided qualitative morphology rather than quantitative three-dimensional measurements of mesh dimensions, connectivity, or tortuosity. Water interactions with individual proteins, polysaccharides, and interfaces were not measured directly, and WAI remains empirical. Future studies should integrate TD-NMR diffusion with confined compression, poroelastic or permeability analysis, quantitative three-dimensional imaging, mastication-induced serum collection, tribology, and dynamic sensory measurements. A further limitation is that sensory ratings and TD-NMR measurements were obtained from separate experimental units, precluding a valid within-sample correlation between D and SJI in the present dataset. Conclusions Within the two formulations examined, bulk moisture retention did not parallel the observed sensory juiciness contrast. Initial moisture content did not differ significantly between formulations, and the soy-based burgers showed lower cooking loss but substantially lower sensory juiciness. In contrast, TD-NMR water self-diffusion clearly differentiated the two matrices at the parent-burger level, with higher D in the beef formulation that was also rated as juicier by the sensory panel. Because these measurements were obtained from separate experimental units, this result represents concordant formulation-level evidence rather than a within-sample correlation. The interpretation is deliberately multiscale. Molecular self-diffusion is not equivalent to serum-release velocity during mastication. D characterizes spontaneous translational molecular motion in the pre-deformation matrix, whereas mastication generates deformation and pressure gradients that drive macroscopic fluid displacement. The present results therefore support D as a sensitive non-destructive descriptor of the pre-deformation water state in matrices with contrasting sensory juiciness, but they do not demonstrate that diffusion itself causes juice release or that D predicts SJI within individual burgers. Potential common structural determinants include network and aqueous-domain dimensions, connectivity, tortuosity, interfaces, and water-matrix interactions. Direct studies measuring TD-NMR, mechanically induced serum release, and sensory juiciness in the same independently prepared samples will be required to establish the proposed causal bridge. CRediT authorship contribution statement Moshe H. Azachi: Methodology, Investigation, Data curation. Zeev Wiesman: Conceptualization, Supervision, Project administration, Writing - original draft, Writing - review & editing. Data availability Data will be made available on reasonable request. Ethics statement The sensory evaluation involved adult trained panelists and was conducted in accordance with applicable institutional guidelines for sensory research involving human participants. The study was determined to be exempt from formal ethics review by Ben-Gurion University of the Negev. All participants were informed of the study procedures and provided informed consent before participation. Participation was voluntary, and sensory data were collected and analyzed anonymously. Declaration of generative AI and AI-assisted technologies in the manuscript preparation process During the preparation and revision of this work, the authors used OpenAI ChatGPT to assist with language editing, organization, and refinement of the manuscript text and scientific presentation. Generative AI was not used to generate experimental measurements or alter the experimental dataset. After using this tool, the authors critically reviewed, verified, and edited the content as needed and take full responsibility for the content of the publication. Funding This research was partially supported by the Goren Fund for biotechnology research at Ben-Gurion University of the Negev. Declaration of competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. This declaration is made by the corresponding author on behalf of all authors for the revised submission of manuscript CRFS-D-26-01700 to Current Research in Food Science. Acknowledgments The authors gratefully acknowledge Dr. Z. Abramovich for professional and technical support and thank the students who contributed to the laboratory work associated with this study. Handling Editor: Professor Alejandro G.Marangoni Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.crfs.2026.101562. Supplementary data The following is the Supplementary data to this article: Multimedia component 1 References Associated Data Supplementary Materials Multimedia component 1 Data Availability Statement Data will be made available on reasonable request.