Demographic information regarding the participants of the 36 studies selected for qualitative synthesis is shown in Table 2. Notably, only three of these studies described swallowing or oral processing in children; one of these was a study of swallowing in premature infants , while a second  explored differences in chewing behaviors in infants aged 6 months to 2 years of age. The third study involving children explored oral processing behaviors in two groups of typically developing girls aged 5 and 8 years old, as well as a control group of healthy adult women . Of the 29 studies describing swallowing or oral processing in adults, 27 reported data for healthy adult participants [19–45], with two studies restricting their focus to denture wearers [46, 47]. A total of 10 studies reported data for adults with dysphagia [20–24, 48–52]. Four of these studies described swallowing in stroke patients in comparison to healthy controls [21–24] and a 5th paper described a group of patients with dysphagia secondary to Chagas’ disease, again with comparison to a group of healthy controls . Two papers reported data for individuals with dysphagia related to head and neck cancer, in one case following surgical resection of the soft palate  and the second exploring post-radiation dysphagia in patients treated for nasopharyngeal carcinoma . The remaining papers described swallowing in patients with Parkinson’s disease , in unspecified neurogenic dysphagia , or in unspecified dysphagia . Sample sizes ranged from 3  to 205  participants.
The various food and liquid stimuli used in the studies selected for inclusion in the qualitative synthesis are summarized in Tables 3 (radio-opaque liquid stimuli), 4 (non-opaque liquid stimuli), and 5 (solid stimuli). Of the 36 studies selected for detailed review, seven reported comparative data for swallows of thin liquid (either barium, water or juice) and an extremely thick liquid (i.e., pureed or spoon-thick consistency) [19, 26, 33, 35, 48, 51, 53]. A total of 13 articles described swallowing measures for a narrower contrast, i.e., thin liquid compared to either a mildly thick liquid (also known as nectar-thick) [17, 19, 23, 26, 28–30, 32, 35, 48, 51, 53] and/or a moderately thick liquid (also known as honey-thick) [28–30, 32, 33, 35, 53], with six of these articles including both mildly thick and moderately thick liquids [28–30, 32, 35, 53]. In terms of solid stimuli, which were explored in a total of 18 studies (Table 5) [18, 19, 21, 24, 27, 31, 39–48, 51, 52], there were effectively no stimuli that were the same in any two or more studies. Solid foods ranged from items that were described by authors as being softer (i.e., banana with barium paste ; cooked rice mixed with barium ; corned beef ; gummy bears ; konjac jelly , or gelatin cubes ) to items at the harder end of the continuum (e.g., fresh raw carrots ; biscuits or cookies [24, 31, 48, 51] or peanuts ). The description of certain items as “softer” in these studies illustrates the subjectivity with which texture descriptors may be applied. For example, “crisp, peeled apple” were described as being “softer”  in comparison to raw carrot , despite the fact that a crisp apple would not generally be regarded as a soft texture. Three studies, all originating from Asia, explored the combination of solid and liquid consistencies using either corned beef in a liquid barium , a thick rice gruel (the consistency of which was not further described)  or 12 g of cooked rice added to 100 ml of liquid barium .
Given the available data, it appears reasonable to synthesize observations regarding differences in swallowing physiology and function across the spectrum of liquid consistencies, from the thin to the extremely thick end of the continuum. However, caution is warranted with respect to delineating quantitative values to capture levels or categories of liquids along this continuum, based on incomplete reporting and the variety of methods and measures used to characterize liquid flow in the studies reviewed. This variety challenges the idea that the stimulus labels used in the literature (e.g., thin, nectar-thick, honey-thick) map to defined ranges of flow. For example, a wide variety of different studies reported using thin liquid barium, but where recipes were reported, these used different concentrations of barium and different dilutions with water or other thin liquids. Insufficient information was provided in the majority of these studies to support recipe replication, or to calculate the weight to volume concentrations of the resulting barium suspensions. Furthermore, given that commercial barium preparations frequently involve additional components to reduce foaming or aid suspension, including gums and starches, viscosity cannot be presumed without additional information.
Very few studies provided objective measures of stimulus characteristics such as viscosity, yield stress, or density (see Tables 3, 4). In several cases, the authors used metaphors to describe the apparent viscosities of stimuli, such as “with a viscosity similar to water”, but failed to provide adequate evidence to support these descriptions. Indeed, several of the metaphors used were scientifically implausible; for example, 120 % w/v E-Z-HD barium is described as being similar in viscosity to water in one study , although barium solutions typically have non-Newtonian flow characteristics and viscosities well above those of water [54, 55]. The fact that both starch- and xanthan-gum thickeners are acknowledged to produce liquids with non-Newtonian flow [54, 56–62] presents a challenge when comparing the stimuli used across these studies; the measured value of viscosity (i.e., “apparent viscosity”) is very sensitive to the shear rate at which the measurement is taken. In cases where viscosity measures were reported, the literature lacked any apparent convention with respect to reporting values at specific shear rates. From the data reported, it can be noted that the non-opaque stimuli labeled as “thin” had viscosities ranging up to 12 mPa s @ 45/s , while the radio-opaque “thin” liquid stimuli spanned a larger viscosity range, reaching reported values as high as 351 mPa s at 25/s . Non-opaque liquids described as mildly thick or nectar-thick had viscosities as high as 466 mPa s at 25/s  or 325 mPa s at 45/s , while the radio-opaque liquids in this category had viscosities up to 863 mPa s at 25/s . Similarly, the stimuli labeled as moderately thick or honey-thick had viscosities reaching 1,541 mPa s at 25/s  for radio-opaque liquids or 785 mPa s at 45/s  for non-opaque stimuli. It is interesting that even among manuscripts arising from the same lab [28, 29] there is no clear convention regarding the shear rates at which viscosities are reported. Shear rate is the term used to describe the rate of deformation of non-Newtonian stimuli as the fluid layers slide over each other when the bolus is placed under stress or force. During swallowing, shear rate for a bolus may be altered by the speed of biomechanical events including tongue movement and pharyngeal shortening or constriction. Perceptual experiments suggest that a range of shear rates is likely to be operational in the mouth during oral processing and swallowing [59, 63, 64]. However, there is no clear guidance from the literature regarding the shear rates that should be used as references when reporting the apparent viscosities of food and fluid stimuli that are being studied. Such variation in reporting makes for confusion and limits generalizability across studies.
Risk of Bias
The evaluation for risk of bias was performed according to the guidelines suggested by the Cochrane Bias Methods Group . Specifically, the methods of each study were reviewed to determine whether there was potential bias in terms of participant selection, the performance of the particular study tasks by the participants, the detection or measurement of behaviors of interest, attrition or missing data, and reporting of results. As shown in Table 6, for the 36 studies reviewed, there were identified risks with respect to bias for every single study. By far, the most common risk of bias lay in the failure to report whether or not raters were blind to bolus consistency during analysis. In some cases, blinding to participant identity was reported, but given the nature of our interest in determining whether there are objective differences in swallowing or oral processing behaviors across boluses with different textures, blinding to stimulus consistency is an important consideration. It may well be that in some cases, such as videofluoroscopy, blinding to bolus consistency is less practical or feasible; however, the literature reviewed lacked acknowledgment of this issue entirely. This may reflect the fact that the primary question in many of these studies was something other than measuring differences in swallowing as a function of bolus consistency; nevertheless, in future studies where this is the purpose, blinding to bolus type would be desirable to limit bias during data analysis. Similarly, in the majority of cases, the reported data appeared to arise from analysis by a single rater with no reporting of inter- or intra-rater reliability. In some cases, measures appeared to be taken online and involved some degree of subjectivity, such that measurement validity and reliability are concerns for many of the studies reviewed. Finally, a subtle but important risk of bias must be mentioned regarding this literature to the extent that investigators selected particular stimuli to study and the reasons guiding these choices were not always reported. As described in the previous section, the stimuli covered by this literature represent a wide variety of discrete points along any sort of viscosity or material characteristic continuum. As such, caution is warranted in drawing conclusions that may be generalized to other ranges on these continua.
Observed Trends and Levels of Evidence
Notwithstanding the caveats mentioned in the previous three sections, the identified studies do provide sufficient preliminary information to support a trend analysis regarding differences in swallowing physiology and function related to differences in stimulus consistency. Table 7 summarizes the main findings from each of the 36 reviewed studies, which are grouped according to the type of instrumentation used to measure swallowing or oral processing behavior. Videofluoroscopy and surface electromyography were used in 12 and 10 studies, respectively, thereby accounting for the bulk of the observed trends, but in total, 12 different types of instrumentation were used.
The level of evidence for each main finding is shown in the far right column of Table 6, according to the scheme used by the National Health and Medical Research Council of Australia . It can be noted that the selected studies fall into one of two types with respect to level of evidence. In total, 28 studies [17, 25–33, 36–53] were classified as reporting level IV evidence, that is, evidence arising from case series, post-test or pre-test and post-test studies without any comparison to controls. The remaining 8 studies [18–24, 35] were classified as level III-2 studies, reporting evidence from non-randomized cohort, case–control or interrupted time-series studies involving comparison to a control group.
Comparing results across studies, it is possible to identify patterns associated with thickened liquids or food texture modification. With respect to liquids, thicker liquids were reported to increase the duration of swallowing events compared to thin liquids in accelerometry , electromagnetic articulography , ultrasound  and surface electromyography signals [19, 26, 30], and also on videofluoroscopy for pharyngeal transit time measures [17, 24]. In patients with stroke-related dysphagia, longer upper esophageal sphincter opening durations  were also reported for paste consistency stimuli than with thin liquids, while longer oral transit times were observed with the paste consistency in patients with Parkinson’s disease  and those who had received radiation therapy for nasopharyngeal carcinoma . Electromyographic measures of oral processing duration were longer for agar gels compared to water data . Findings regarding the influence of liquid consistency on pharyngeal delay times in stroke patients were equivocal, with one study reporting longer delays with a pudding-thick consistency  and a second study reporting the opposite trend . Two further reports found results that conflicted with the generally observed trend of longer duration events being seen with increasing viscosity. One study reported that the sounds associated with swallowing water were longer than those seen with either yogurt or konjac jelly . Hyoid movement durations were also described to be shorter with paste consistency compared to thin barium following radiation treatment for nasopharyngeal carcinoma . Two studies describe measures that did not change as a function of liquid consistency: swallow apnea duration was reported to remain unaffected by bolus consistency in healthy adults  while measures of swallow response time (also known as stage transition duration) and laryngeal vestibule closure duration did not differ for a thin to nectar-thick liquid barium comparison in stroke patients .
In addition to observations regarding physiological timing measures, other reported measures support the impression that thicker and harder items require greater effort in oral processing and swallowing. Measures that contribute to this observation include more prominent and well-defined accelerometry signal peaks , greater variability in tongue movement patterns [28, 29], higher surface electromyography amplitudes [19, 36, 41, 47], higher velocities of jaw movement , greater variability in surface electromyography patterns , and increased amplitudes of tongue-palate pressure . Several studies concur that boluses with increased hardness elicit timing differences in chewing, involving faster rates, longer cycle durations, and a greater number of cycles [18, 41–44, 46, 47]. Findings were mixed with respect to the influence of bolus consistency on the magnitude of hyoid and laryngeal movements. One large study reported larger hyoid and laryngeal excursion for paste consistency and bread boluses compared to thin liquids , while a smaller study in healthy adults failed to find differences across different solid boluses .
With respect to functional swallowing measures, an important question is to determine whether penetration–aspiration of material into the airway is effectively reduced by altering bolus consistency? Several of the videofluoroscopic studies concur on this question, as illustrated in Fig. 2a, b. Bingjie and colleagues reported that the frequency of penetration–aspiration in stroke patients decreased as liquid viscosity increased . This trend was also seen in the studies by Chen et al. , Barata et al. , Troche et al.  and by Lee et al. , who further described that aspiration was worst for thin liquids, better with a mixed consistency involving rice in liquid barium, and best for rice served without combining it with liquid. However, a cautionary note is also warranted on the basis of the selected studies, in that greater vallecular residue was observed with paste consistency barium than with thin liquid barium [48, 49] and with a plain rice bolus compared to a rice and barium mixed consistency . Troche and colleagues  also observed that patients with Parkinson’s disease used a greater number of tongue pumps to successfully swallow a pudding-thick consistency, than for a thin liquid bolus, suggesting that clearance was worse with the thicker consistency. A recent report by Hind and colleagues , also reports a trend toward greater pharyngeal residues for barium stimuli with increasing viscosity.
An interesting study exploring swallowing with liquid barium, mixed consistency and a solid food (corned beef) demonstrated that for mixed consistencies and the solid food, the location of the bolus at swallow onset was lower in the hypopharynx than with liquids . However, in a clever twist in their experimental design, these authors also asked participants to engage in chewing with the liquid barium stimulus and showed that this led to accumulation of the liquid bolus in the vallecular space, as seen with the mixed and solid consistencies.
With respect to solid foods, the literature search identified several studies in which rheological or texture profile analysis methods were used to measure the characteristics of the food bolus were at the end of oral processing, when the bolus was considered to be ready for swallowing [21, 40, 42–44]. These articles suggest that the property of cohesiveness remains stable during chewing and oral processing while other mechanical properties change  and are influenced by dry matter content , the degree to which salivary enzymes are absorbed by the bolus and contribute to starch hydrolysis , and the composition of the bolus with respect to the use of single gelling agents or complex gel combinations [43, 44]. In the food oral processing literature, the construct of cohesiveness is defined as a mechanical textural attribute relating to the degree to which a substance can be deformed before it breaks. The standard method for measuring cohesiveness during sensory panel testing involves placing a sample between the molar teeth, compressing the sample, and evaluating the degree of deformation before rupture [68–70]. Adjectives that are listed as descriptors of cohesiveness include: fracturable, crumbly, crunchy, brittle, crispy, crusty, chewy, tender, tough, short, mealy, pasty, and gummy. The texture reference scale developed by Munoz  is recommended in ISO guidelines for sensory ratings of cohesiveness, but it is acknowledged that no suitable set of reference products has been developed for this attribute.