**1. Introduction**

The sensory methodology for virgin olive oils (VOOs) known as the "panel test" was proposed in 1987 [1] and, to date, represents the most valuable approach to assess sensory characteristics and quality for consumer and producer protection [2]. The purpose of the method is to standardize procedures for evaluation of the organoleptic characteristics of VOOs and to establish specific quality grades (extra virgin olive oil—EV, virgin olive oil—V, ordinary virgin olive oil—O, lampante olive oil—L). A group of assessors selected in a controlled manner, suitably trained to identify and measure the intensity of positive and negative sensations, represents the analytic tool of this methodology. A collection of methods and standards has been adopted by the International Olive Council (IOC) for sensory analysis of VOOs. These documents describe the vocabulary that tasters must adopt, the characteristics that the sensory laboratory must possess, the tasting conditions and characteristics of the glass for organoleptic analysis of oils, and the sensory method and rules for the selection, training, and monitoring of skilled virgin olive oil tasters [3–7].

In 1991, the method was included into European regulations and obtained the legal validity for establishing the quality grade of the product that included only three categories of VOOs: EV, V, and L [8]. Since application of the method showed some drawbacks, drawing on its own experience, the IOC made a series of revisions to render the method simpler and more reliable [9].

In 2002, the most important innovation introduced was the application of a statistical index to classify oils according to the median of the main perceived defect (mpd) and the median of the fruity attribute that represents the most important positive descriptor. A limit for the value of the robust variation coefficient, which must be no greater than 20%, was also established. In fact, the use of statistical procedures to analyze sensory data is fundamental, as it provides reliable results that are required for data from other analytical methods [10]. Subsequent amendments and revisions concerned, for example, a list of sensory defects, the specific optional terminology for labeling purposes or tasting conditions, have been adopted up to now [9].

Although it has been responsible for improving the quality of VOOs in the last 28 years, the Panel test is frequently under scrutiny. The problem mainly focuses on the debated classification of borderline oils (EV vs. V, V vs. L), reproducibility of results among different laboratories, the limited number of samples that can be analyzed per day (four samples in each session with a maximum of three sessions per day) and the presence of at least 8–12 trained individuals for each sensory evaluation [6]. Another problem in applying the panel test method depends on the lack of appropriate reference standards for training assessors [11]. In addition, some recent commentaries [12,13] discussed methodological features that should be performed more accurately in order to avoid disagreements between different panels. To overcome this, some strategies were proposed that should be applied during different steps of training of assessors (determination of the group detection threshold, selective trials) or during official tasting sessions (alternative approach to the CVr%) for overcoming these difficulties.

Organoleptic assessment is both a qualitative and quantitative method since its application results in the classification of samples based on the median of the main predominant defect and the presence or absence of the fruity attribute. Consequently, assessors in each panel must be effectively trained for correct classification of samples and for correct recognition of the intensities of perceived attributes.

In this context, the OLEUM project "Advanced Solutions for Assuring Authenticity and Quality of Olive Oil at Global Scale" funded by the European Commission within the Horizon 2020 Programme (2014–2020, grant agreement No. 635690), is engaged in reinforcing the methodology for sensory evaluation through design of a global procedure named the "quantitative panel test". This approach aims to improve the activity of sensory panels, whose work remains central to ensuring the quality of the product by: (i) reducing the number of samples to be assessed by the sensory panel by establishing chemometric models (calibrated on a large dataset of reliable sensory classified VOOs) that are able to predict assignment of samples to a specific quality grade using rapid instrumental screening methods, which could allow pre-classification with a certain level of probability that can allow the panels to focus more on sensory analysis of uncertain samples; (ii) increasing the panel's performance by introducing new artificial reference materials validated by a number of sensory panels (six in the case of the OLEUM project) and formulated ad hoc to resemble specific sensory attributes (e.g., rancid and winey); (iii) relating attributes and defects found in VOO with specific molecules (volatile compounds) in order to have an additional qualitative and quantitative tool (quantitation of specific volatile compounds) to support the panel test in confirmatory analyses or in cases of disagreement between panels.

In this regard, some recent works deal with the monitoring of the presence of molecular markers related to specific sensory defects in VOO headspace [14] together with the setting up of chemometric models based on volatile compounds for the prediction of sensory characteristics [15].

The present paper does not aim to illustrate the entire scheme and all the methods involved in the "quantitative panel test". However, in the framework of the panel test, it highlights possibilities for amelioration and describes the proficiency improvement given by formative training, and the method used to obtain sensory classified samples from analysis of a set of VOOs to be used for calibration of rapid instrumental screening methods. Herein, the results of the sensory evaluation of 334 samples are reported and discussed with the aim to: (i) verify the effectiveness of application of the sensory method to evaluate the quality of the product according to [8] and latter modifications (EV, V, L); (ii) highlight the importance of the strict application of IOC guidelines for quality control methodology (for selecting, training, and monitoring tasters and panels) by the sensory panels in performing organoleptic assessment of VOO; (iii) obtain the most reliable sensory classification of samples in terms of quality grades and intensities of positive (fruity) and negative (main perceived defect/s) attributes; this can be achieved by application of a newly proposed "decision tree" that includes formative reassessments when misalignments on the category or the main perceived defect and/or fruity attribute occur.

Many studies in the literature have discussed the relationship between the official sensory method applied by a trained panel and consumer perception [16–21], but to our knowledge there are few studies comparing the results of different trained panels and none aimed to reinforce the application of the official method and increase harmonization among different panels. Specifically, the key elements of this work are: (i) the very large dataset obtained by collecting 334 oils from two olive harvest seasons, representative of the most common olive cultivars, different geographical origins, different sensory profiles, and, especially, the main sensory defects perceived; (ii) the processing of data provided by several panels to obtain a reliable classification by the application of a new decision tree useful for possible correlations with instrumental data and/or for building discriminating models by different instrumental approaches.

## **2. Materials and Methods**

## *2.1. Sensory Panels*

Six panels from six different countries were involved in the sensory analysis carried out in the OLEUM project: EUROFINS from Germany, coded as EU; IPTPO from Croatia, coded as IP; ITERG from France, coded as IT; UNIBO from Italy, coded as UN; UP/ZRS from Slovenia, coded as UP/ZRS; and UZZK from Turkey, coded as UZ.

Each panel has some sort of public authority recognition (national authorities; International Olive Council, IOC; national accreditation bodies for EU standards) [22] and takes part in national and international interlaboratory proficiency tests (organized by private or public authorities) and/or IOC interlaboratory comparison. Their sensory activities are focused on evaluation of the grade of quality (quality control), PDO/PGI certification, olive oil competition, and sensory analysis of samples involved in scientific research. The number of samples evaluated each year by the six panels varies from 125 to 1800. The UNIBO panel was responsible for coordinating the activities of panels and for elaboration of sensory data.
