Fresh de-boned chicken breast was used throughout this study. Samples were purchased from a local supplier. The stock was stored in a freezer at −20˚C until used. It was then thawed in a refrigerator at 4˚C for 24 to 36 h before commencement of the experiment.

The thawed chicken breast meat was cut into approximately 10 × 20 × 15 mm slabs. The sample slabs were individually deep-fat fried in a liquid shortening (Can Amera Food, Oakville, ON) in a commercial programmable computerized pressure deep-fat fryer (Henny Penny Computron 7000 Pressure Fryer, Model 500C, Henny Penny Corporation, Eaton, OH). The fryer was preheated at 170˚C for 2 h prior to the commencement of frying. Samples were placed in a wire basket to ensure good contact between the sample and the oil. Fried samples were cooled at room temperature, weighed, and placed in plastic Zip-Lock sample bags until the density measurements were to be executed.

The moisture content was determined at the end of each frying period. Moisture content was determined based on AOAC Method 934.01 [26] . Two grams of fresh samples were placed in an oven at 100˚C for 18 h and the final moisture content was computed based on weight differences. In preparation of fat analysis the samples were freeze fried using a freeze-dryer ModulyOD-115 (Thermo Savant, Holbrook, NY) for 30 h. Prior to freeze drying, all samples were frozen overnight in a freezer at −20˚C. The freeze-dryer’s refrigerating system and vacuum chamber were operated at −50˚C and 750 mm Hg, respectively. This approach was adopted to minimize major post changes in the sample’s microstructure [25] .

The fried, freeze-dried samples were ground using a blender (Proctor-Silex, model E160B, Picton, ON), and fat was extracted with a solvent extractor (SER148, Velp Scientifica, Usmate, Italy) using petroleum ether based on AOAC Methods 945.16 (AOAC, 2007). Ground samples were weighed and placed in thimbles (procedure followed as recommended by (SER148, Operation manual) for details). The oil content (OC) dry basis was computed for each sample based on the following relationship (Equation (1)).

Bulk density: Bulk or total density of the samples was measured using the liquid displacement technique with water. The method requires coating sample with wax (to ensure no liquid penetration of the samples during the test) and immerse in a volumetric displacement (Archimedes) pycnometer. The apparatus was filled with water and the samples were immersed in the sample compartment, which was then hermetically sealed with the lid. The volume displacement was measured by turning the apparatus upside down twice, the first time without the samples and the second time with the samples immersed in the liquid. The volume displacement was computed based on the follow relationship (Equation (2)).

where V_s is the volume of the samples, V_f is the final volume of the samples immersed in the liquid, and V_i is the initial volume without the samples, the volume of the wax was subtracted from the final volume. The shrinkage volume was measured using the same method.

Apparent density: The apparent or skeletal density was measured using a helium pycnometer (Model 1305 Multivolume, Micromeritics Instrument Corporation, Norcross, GA). The fried samples were weighed prior to analysis. The solid volume of the samples excludes pores within the sample material. Using the pycnometry method, the sample was placed in the 35 cm³ sample chamber and was subjected to cyclic action (purging) by a pressurizing and depressurizing with helium gas prior to analysis in order to expel all the air and vapors trapped in the pores and crevices. The analysis was conducted at a room temperature with pressure of 134 ± 1.4 kPa (19.5 ± 0.2 psi). Initially, all valves were closed while the system equilibrated to the atmospheric pressure. The inlet valve was opened to fill the sample chamber with helium at an elevated pressure and closed for 15 s of pressure equilibrium, after which time the pressure was recorded as P₁. Then intermediate valve was opened, allowing the trapped helium to flow into the expansion chamber, and pressure P₂ was measured. The exhaust valve was then opened to release the spent helium gas from the system. The following general equation (Equation (3)) was used to compute the volume of the samples. For detailed methodology and procedure, refer to the standard protocol [27] .

where: V_Samp = the volume of the samples of interest; V_Cell = volume of the sample cell with the empty sample cup in place; V_Exp = the volume of the expansion chamber; P₁ = sample chamber initial pressure with the expansion chamber closed; and P₂ = final chamber pressure with the expansion chamber open.

Density is a physical property of all matter; it is simply the unit quantity of matter per volume of the same quantity (kg∙m⁻³). The densities used in this study are defined as follows:

Bulk density: the measure of bulk density includes all pores, interparticle spaces, moisture, and air in the material.

Apparent density: this measure includes blind and non-interconnected pores of the material and excludes open, interconnected and interparticle pore spaces.

A full (3 × 20) factorial design with 3 replications was used in this study. The factors were frying oil temperatures (FOT) with three levels (170˚C, 180˚C, and 190˚C), and frying time with 20 levels, and samples were fried randomly at each 20 frying times (5, 10, 15, 20, 30, 45, 60, 90, 120, 150, 180, 210, 240, 300, 360, 420, 480, 540, 600, and 900 s)within each FOT. The analysis of variance (ANOVA) and non-linear regression were conducted using SAS software (SAS 2012). All mean comparisons were performed by using Duncan’s multiple range test (DMRT), at the 5% level of significance.

Bulk and particle densities of deep-fat fried de-boned chicken meat samples are shown in Figure 1 and Figure 2, respectively. The bulk density decreased with frying time, whereas the apparent density increased with frying time, Wang and Brennan [28] ) reported similar trends in air-drying of potato. The analyses of variance (ANOVA) of the two densities show that the temporal changes (P < 0.01) were influenced by frying oil temperature (FOT). Synergy between FOT and frying time significantly (P < 0.05) impacted the changes in density.

The mean apparent density was not significant different at FOT 180˚C and 190˚C. The bulk density values decrease from 1.15 g/cm³ (control, for which there was no moisture loss) to 0.98, 0.95, and 0.92 g/cm³ after 900 s of frying at the following FOTs 170˚C, 180˚C and 190˚C and final moisture contents of 33%, 25%, and 16% wb, respectively. Frying at a higher temperature resulted in a lower bulk density. This is due to the effect of aggravated water loss resulting in pore formation and subsequent pore structure collapse resulting to shrinkage, was also observed by Rahman et al. [12] in drying fish muscles. In contrast, the apparent density increased from 1.13 g/cm³ (for the control) to 1.25, 1.24 and 1.22 g/cm³. The apparent density increased consistently when fried for 540 s, and the mean values were 1.21, 1.29 and 1.23 g/cm³ at 170˚C, 180˚C and 190˚C, respectively, after which it tended to equilibrate. This behavior was analogous to the oil absorption curves during deep-fat frying reported by Kassama and Ngadi [23] . The morphological change that modulates densities variation could also be attributed to physicochemical reaction induced by the deterioration of solid content and collageneous connective tissues. Denaturation of protein is also a significant factor that contributes to alteration of the physical properties, thus change in densities. Additionally, the change in state could also be attributed to oil absorbed into the empty voids, hence the volumetric gain due to oil uptake in place of voids evacuated by moisture may explain the decreasing trend of the apparent density Kassama and Ngadi [3] , Krokida et al. [29] made similar observations for deep-fat fried potato strips.

Figures 3(a)-(c) show bulk and apparent densities as functions of moisture loss during deep-fat frying. Bulk density decreases while apparent density increases, as moisture loss increases. These results agree with the observations of various researchers [19] [22] [30] . The relationships between distinct parameters such as the moisture content, bulk and apparent density and the process variables are shown in the Pearson correlation coefficients matrix (Table 1). A weak linear correlation (r = 0.66; P < 0.01) exists between moisture content and bulk density.

The apparent density peaked at the low moisture region, while the bulk density peaked at the high moisture region. There was a linear relationship (r = 0.83; P < 0.01) between moisture loss and apparent density. The FOT was poorly correlated to both the bulk and apparent density, while frying time was high positively correlated. A weak correlation (r = −0.66; P < 0.01) was achieved between the bulk and apparent density of de-boned chicken meat during deep-fat frying. Therefore, the bulk and apparent density may be adequately predicted by a

P < 0.01.

linear equation with a higher degree of accuracy as a function of moisture loss in de-boned chicken meat during the deep-fat frying process. Balaban and Pigott [18] reported a best fit with the R²-value of >0.86 for density for fish muscle during drying by assuming linearity in their application of the drying model equation which will be appropriate to our study.

The experimental data were fitted to the following empirical equations (Equation (4) and Equation (5)) to predict the bulk and apparent density of chicken meat during deep-fat frying.

where ρ_B is the bulk density, ρ_oB is the initial bulk density, ρ_A is the apparent density, ρ_oA is the initial density, a is a constants and X is the moisture loss.

Table 2 shows the corresponding values of the constants and the coefficient of determination (R²) obtained from a linear regression analysis. The estimation of density parameters by Equation (4) and Equation (5) shows a good fit, and thus the coefficients of the determinations are relatively high for all ranges of the FOT as seen in Figure 4 the experimental and predicted curves as a function of moisture loss.

Shrinkage is a change in volume induced by moisture loss and probably formation and collapse of the pore structures. Shrinkage in meat is anisotropic [3] [31] . Rahman and Perera [32] reported that the rate of shrinkage is dependent in the muscle fiber orientation. In biological materials, the rate of shrinkage is dependent on orientation of the fiber stretch either in the longitudinal or transversal plane during frying [22] [33] . Shrinkage is temperature-dependent and excessive shrinkage causes distortion of shapes, hence affects the final texture and quality attributes of a food product. The ANOVA for volumetric shrinkage data shows significant changes (P < 0.05) in time. The trend in time was not affected by temperature as indicated by the non-significant interaction.

Shrinkage increases logarithmically with time as seen by the shape of the curve in Figure 5. The rates of shrinkage were 0.013, 0.001, and 0.008/s and occurred rapidly during the first 90 s of FOT 170˚C, 180˚C, and 190˚C, respectively and decreased as frying time increased. Thus, there was a linear relationship between shrinkage and moisture loss as shown in Figure 6. However, the rate of volumetric shrinkage slows down after 300 s of frying probably due to the oil uptake at low moisture regions. A linear relationship was observed at all levels of FOT, similar correlations were reported by others researchers on drying of foods [17] [19] [23] [30] . The regression equations for each temperature are given in Table 3 and the correlation coefficients indicate a close fit to the experimental data.